In-home service robots embodied as mobile platforms with onboard cameras are increasingly being proposed for well-being monitoring and fall detection for older adults. Yet, how users perceive a robot’s movement and observation behavior within the home remains underexplored. This work examines user preferences for robot-based observation strategies in Human Activity Recognition (HAR) and evaluates how these strategies affect recognition performance. In a within-subject study with adults over 50, we compare stationary versus adaptive-distance observation behaviors. Results reveal that while stationary observation is generally preferred for being less intrusive, preferences vary depending on the activity context. HAR accuracy remains comparable across both strategies, and combining robot and ambient sensing enhances recognition of complex, temporally extended activities.

Ambiguities are inevitable in human-robot interaction, especially when a robot follows user instructions in a large, shared space. For example, if a user asks the robot to find an object in a home environment with underspecified instructions, the object could be in multiple locations depending on missing factors. For instance, a bowl might be in the kitchen cabinet or on the dining room table, depending on whether it is clean or dirty, full or empty, and the presence of other objects around it. Previous works on object search have assumed that the queried object is immediately visible to the robot or have predicted object locations using one-shot inferences, which are likely to fail for ambiguous or partially understood instructions. This paper focuses on these gaps and presents a novel model-agnostic approach leveraging semantically driven clarifications to enhance the robot’s ability to locate queried objects in fewer attempts. Specifically, we leverage different knowledge embedding models, and when ambiguities arise, we propose an informative clarification method, which follows an iterative prediction process. The user experiment evaluation of our method shows that our approach is applicable to different custom semantic encoders as well as LLMs, and informative clarifications improve performances, enabling the robot to locate objects on its first attempts. The user experiment data is publicly available at https://github.com/IrmakDogan/ExpressionDataset.

Object rearrangement is a key task for household robots requiring personalization without explicit instructions, meaningful object placement in environments occupied with objects, and generalization to unseen objects and new environments. To facilitate research addressing these challenges, we introduce PARSEC, an object rearrangement benchmark for learning user organizational preferences from observed scene context to place objects in a partially arranged environment. PARSEC is built upon a novel dataset of 110K rearrangement examples crowdsourced from 72 users, featuring 93 object categories and 15 environments. To better align with real-world organizational habits, we propose ContextSortLM, an LLM-based personalized rearrangement model that handles flexible user preferences by explicitly accounting for objects with multiple valid placement locations when placing items in partially arranged environments. We evaluate ContextSortLM and existing personalized rearrangement approaches on the PARSEC benchmark and complement these findings with a crowdsourced evaluation of 108 online raters ranking model predictions based on alignment with user preferences. Our results indicate that personalized rearrangement models lever-aging multiple scene context sources perform better than models relying on a single context source. Moreover, ContextSortLM outperforms other models in placing objects to replicate the target user’s arrangement and ranks among the top two in all three environment categories, as rated by online evaluators. Importantly, our evaluation highlights challenges associated with modeling environment semantics across different environment categories and provides recommendations for future work.

Collaborative human-robot task execution approaches require mutual adaptation, allowing both the human and robot partners to take active roles in action selection and role assignment to achieve a single shared goal. Prior works have utilized a leader-follower paradigm in which either agent must follow the actions specified by the other agent. We introduce the User-aware Hierarchical Task Planning (UHTP) framework, a communication-free human-robot collaborative approach for adaptive execution of multi-step tasks that moves beyond the leader-follower paradigm. Specifically, our approach enables the robot to observe the human, perform actions that support the human’s decisions, and actively select actions that maximize the expected efficiency of the collaborative task. In turn, the human chooses actions based on their observation of the task and the robot, without being dictated by a scheduler or the robot. We evaluate UHTP both in simulation and in a human subjects experiment of a collaborative drill assembly task. Our results show that UHTP achieves more efficient task plans and shorter task completion times than non-adaptive baselines across a wide range of human behaviors, that interacting with a UHTP-controlled robot reduces the human’s cognitive workload, and that humans prefer to work with our adaptive robot over a fixed-policy alternative.

We introduce a novel feature-based active learning approach that leverages leading feature selection methods from explainable AI (XAI) to elicit user corrections on both the classification label and the leading features contributing to the classification decision. Uniquely, our approach can handle such user corrections, which identify missing and superfluous features in a model-agnostic manner that can be applied to any supervised classification model. Our approach generates augmented samples in a fine-grained manner according to user corrections, and then adds them to the training data to teach relevant and irrelevant features. Our results show that our approach outperforms traditional active learning and a leading baseline in two domains, which includes improvement in classification performance from 0.572 to 0.637 given the same number of queries.

Creative Problem Solving (CPS) is a sub-area within artificial intelligence that focuses on methods for solving off-nominal, or anomalous problems in autonomous systems. Despite many advancements in planning and learning in AI, resolving novel problems or adapting existing knowledge to a new context, especially in cases where the environment may change in unpredictable ways, remains a challenge. To stimulate further research in CPS, we contribute a definition and a framework of CPS, which we use to categorize existing AI methods in this field. We conclude our survey with open research questions, and suggested future directions.

We posit that predicting sensor event sequence (SES) in a smart home can proactively support resident activities or recognize activities that have not been completed as intended and alert the resident. To realize this application, we propose a framework to support accurate SES prediction by leveraging online activity recognition. Our framework includes a novel method of applying a GPT2-based model, which is a sentence generation model, for SES prediction by taking advantage of the property that the relationship between ongoing activity and SES patterns is similar to the relationship between topic and word sequence patterns in NLP. We evaluated our method empirically using two real-world datasets where residents perform their usual daily activities. Our experimental results show the use of the GPT2-based model significantly improves the F1 value of SES prediction from 0.461 to 0.708 compared to the state-of-the-art method, and that using ongoing activity can further improve performance to 0.837. We found that the performance of the online activity recognition model required to achieve these SES predictions was about 80%, which could be achieved using simple feature engineering and modeling.

Object rearrangement is the problem of enabling a robot to identify the correct object placement in a complex environment. Prior work on object rearrangement has explored a diverse set of techniques for following user instructions to achieve some desired goal state. Logical predicates, images of the goal scene, and natural language descriptions have all been used to instruct a robot in how to arrange objects. In this work, we argue that burdening the user with specifying goal scenes is not necessary in partially-arranged environments, such as common household settings. Instead, we show that contextual cues from partially arranged scenes (i.e., the placement of some number of pre-arranged objects in the environment) provide sufficient context to enable robots to perform object rearrangement without any explicit user goal specification. We introduce ConSOR, a Context-aware Semantic Object Rearrangement framework that utilizes contextual cues from a partially arranged initial state of the environment to complete the arrangement of new objects, without explicit goal specification from the user. We demonstrate that ConSOR strongly outperforms two baselines in generalizing to novel object arrangements and unseen object categories. The code and data are available at https://github.com/kartikvrama/consor.

Intelligent decision support (IDS) systems leverage artificial intelligence techniques to generate recommendations that guide human users through the decision making phases of a task. However, a key challenge is that IDS systems are not perfect, and in complex real-world scenarios may produce suboptimal output or fail to work altogether. The field of explainable AI (XAI) has sought to develop techniques that improve the interpretability of black-box systems. While most XAI work has focused on single-classification tasks, the subfield of explainable AI planning (XAIP) has sought to develop techniques that make sequential decision making AI systems explainable to domain experts. Critically, prior work in applying XAIP techniques to IDS systems has assumed that the plan being proposed by the planner is always optimal, and therefore the action or plan being recommended as decision support to the user is always optimal. In this work, we examine novice user interactions with a non-robust IDS system – one that occasionally recommends suboptimal actions, and one that may become unavailable after users have become accustomed to its guidance. We introduce a new explanation type, subgoal-based explanations, for plan-based IDS systems, that supplements traditional IDS output with information about the subgoal toward which the recommended action would contribute. We demonstrate that subgoal-based explanations lead to improved user task performance in the presence of IDS recommendations, improve user ability to distinguish optimal and suboptimal IDS recommendations, and are preferred by users. Additionally, we demonstrate that subgoal-based explanations enable more robust user performance in the case of IDS failure, showing the significant benefit of training users for an underlying task with subgoal-based explanations.

As more non-AI experts use complex AI systems for daily tasks, there has been an increasing effort to develop methods that produce explanations of AI decision making that are understandable by non-AI experts. Towards this effort, leveraging higher-level concepts and producing concept-based explanations have become a popular method. Most concept-based explanations have been developed for classification techniques, and we posit that the few existing methods for sequential decision making are limited in scope. In this work, we first contribute a desiderata for defining ``concepts'' in sequential decision making settings. Additionally, inspired by the Protege Effect which states explaining knowledge often reinforces one's self-learning, we explore how concept-based explanations of an RL agent's decision making can in turn improve the agent's learning rate, as well as improve end-user understanding of the agent's decision making. To this end, we contribute a unified framework, State2Explanation (S2E), that involves learning a joint embedding model between state-action pairs and concept-based explanations, and leveraging such learned model to both (1) inform reward shaping during an agent's training, and (2) provide explanations to end-users at deployment for improved task performance. Our experimental validations, in Connect 4 and Lunar Lander, demonstrate the success of S2E in providing a dual-benefit, successfully informing reward shaping and improving agent learning rate, as well as significantly improving end user task performance at deployment time.

Close-proximity human-robot interactions can be improved through the optimization of task-centric factors or by prioritizing the user experience. Prior work has often explored these factors individually. In this paper, we conducted a within-subject study with 18 participants that compared a multi-objective robot motion adaptation method (CoMOTO) against methods that optimize distance from the user (UserAvoidant) or task performance (ShortestPath) in a close-proximity human-robot interaction task. In the task, the robot and participants worked on different tasks in an overlapping workspace. We show that while CoMOTO trajectories took a longer time, they caused significantly fewer interruptions compared to ShortestPath and generated shorter trajectories than UserAvoidant. CoMOTO was also perceived as significantly more intelligent, more trustworthy, and preferred by an overwhelming majority of the participants.

Smart home environments are designed to provide services that help improve the quality of life for the occupant via a variety of sensors and actuators installed throughout the space. Many automated actions taken by a smart home are governed by the output of an underlying activity recognition system. However, activity recognition systems may not be perfectly accurate, and therefore inconsistencies in smart home operations can lead users reliant on smart home predictions to wonder “Why did the smart home do that?” In this work, we build on insights from Explainable Artificial Intelligence (XAI) techniques and introduce an explainable activity recognition framework in which we leverage leading XAI methods (Local Interpretable Model-agnostic Explanations, SHapley Additive exPlanations (SHAP), Anchors) to generate natural language explanations that explain what about an activity led to the given classification. We evaluate our framework in the context of a commonly targeted smart home scenario: autonomous remote caregiver monitoring for individuals who are living alone or need assistance. Within the context of remote caregiver monitoring, we perform a two-step evaluation: (a) utilize Machine Learning experts to assess the sensibility of explanations and (b) recruit non-experts in two user remote caregiver monitoring scenarios, synchronous and asynchronous, to assess the effectiveness of explanations generated via our framework. Our results show that the XAI approach, SHAP, has a 92% success rate in generating sensible explanations. Moreover, in 83% of sampled scenarios users preferred natural language explanations over a simple activity label, underscoring the need for explainable activity recognition systems. Finally, we show that explanations generated by some XAI methods can lead users to lose confidence in the accuracy of the underlying activity recognition model, while others lead users to gain confidence. Taking all studied factors into consideration, we make a recommendation regarding which existing XAI method leads to the best performance in the domain of smart home automation and discuss a range of topics for future work to further improve explainable activity recognition.

Learned knowledge graph representations supporting robots contain a wealth of domain knowledge that drives robot behavior. However, there does not exist an inference reconciliation framework that expresses how a knowledge graph representation affects a robot's sequential decision making. We use a pedagogical approach to explain the inferences of a learned, black-box knowledge graph representation, a knowledge graph embedding. Our interpretable model uses a decision tree classifier to locally approximate the predictions of the black-box model and provides natural language explanations interpretable by non-experts. Results from our algorithmic evaluation affirm our model design choices, and the results of our user studies with non-experts support the need for the proposed inference reconciliation framework. Critically, results from our simulated robot evaluation indicate that our explanations enable non-experts to correct erratic robot behaviors due to nonsensical beliefs within the black-box.

Mixed human-robot teams (HRTs) have the potential to perform complex tasks by leveraging diverse and complementary capabilities within the team. However, assigning humans to operator roles in HRTs is challenging due to the significant variation in user capabilities. While much of prior work in role assignment treats humans as interchangeable (either generally or within a category), we investigate the utility of personalized models of operator capabilities based in relevant human factors in an effort to improve overall team performance. We call this approach individualized role assignment (IRA) and provide a formal definition. A key challenge for IRA is associated with the fact that factors that affect human performance are not static (e.g., one’s ability to track multiple objects can change during or between tasks). Instead of relying on time-consuming and highly-intrusive measurements taken during the execution of tasks, we propose the use of short cognitive tests, taken before engaging in human-robot tasks, and predictive models of individual performance to perform IRA. Results from a comprehensive user study conclusively demonstrate that IRA leads to significantly better team performance than a baseline method that assumes human operators are interchangeable, even when we control for the influence of the robots’ performance. Further, our results point to the possibility that such relative benefits of IRA will increase as the number of operators (i.e., choices) increase for a fixed number of tasks.

Creative Problem Solving (CPS) is a sub-area within Artificial Intelligence (AI) that focuses on methods for solving off-nominal, or anomalous problems in autonomous systems. Despite many advancements in planning and learning, resolving novel problems or adapting existing knowledge to a new context, especially in cases where the environment may change in unpredictable ways post deployment, remains a limiting factor in the safe and useful integration of intelligent systems. The emergence of increasingly autonomous systems dictates the necessity for AI agents to deal with environmental uncertainty through creativity. To stimulate further research in CPS, we present a definition and a framework of CPS, which we adopt to categorize existing AI methods in this field. Our framework consists of four main components of a CPS problem, namely, 1) problem formulation, 2) knowledge representation, 3) method of knowledge manipulation, and 4) method of evaluation. We conclude our survey with open research questions, and suggested directions for the future.

With the growing capabilities of intelligent systems, the integration of robots in our everyday life is increasing. However, when interacting in such complex human environments, the occasional failure of robotic systems is inevitable. The field of explainable AI has sought to make complex-decision making systems more interpretable but most existing techniques target domain experts. On the contrary, in many failure cases, robots will require recovery assistance from non-expert users. In this work, we introduce a new type of explanation, εerr, that explains the cause of an unexpected failure during an agent's plan execution to non-experts. In order for error explanations to be meaningful, we investigate what types of information within a set of hand-scripted explanations are most helpful to non-experts for failure and solution identification. Additionally, we investigate how such explanations can be autonomously generated, extending an existing encoder-decoder model, and generalized across environments. We investigate such questions in the context of a robot performing a pick-and-place manipulation task in the home environment. Our results show that explanations capturing the context of a failure and history of past actions, are the most effective for failure and solution identification among non-experts. Furthermore, through a second user evaluation, we verify that our model-generated explanations can generalize to an unseen office environment, and are just as effective as the hand-scripted explanations.

Requiring multiple demonstrations of a task plan presents a burden to end-users of robots. However, robustly executing tasks plans from a single end-user demonstration is an ongoing challenge in robotics. We address the problem of one-shot task execution, in which a robot must generalize a single demonstration or prototypical example of a task plan to a new execution environment. Our approach integrates task plans with domain knowledge to infer task plan constituents for new execution environments. Our experimental evaluations show that our knowledge representation makes more relevant generalizations that result in significantly higher success rates over tested baselines. We validated the approach on a physical platform, which resulted in the successful generalization of initial task plans to 38 of 50 execution environments with errors resulting from autonomous robot operation included.

When interacting in unstructured human environments, occasional robot failures are inevitable. When such failures occur, everyday people, rather than trained technicians, will be the first to respond. Existing natural language explanations hand-annotate contextual information from an environment to help everyday people understand robot failures. However, this methodology lacks generalizability and scalability. In our work, we introduce a more generalizable semantic explanation framework. Our framework autonomously captures the semantic information in a scene to produce semantically descriptive explanations for everyday users. To generate failure-focused explanations that are semantically grounded, we lever-ages both semantic scene graphs to extract spatial relations and object attributes from an environment, as well as pairwise ranking. Our results show that these semantically descriptive explanations significantly improve everyday users’ ability to both identify failures and provide assistance for recovery than the existing state-of-the-art context-based explanations.

To realize effective heterogeneous multi-robot teams, researchers must leverage individual robots’ relative strengths and coordinate their individual behaviors. Specifically, heterogeneous multi-robot systems must answer three important questions: who (task allocation), when (scheduling), and how (motion planning). While specific variants of each of these problems are known to be NP-Hard, their interdependence only exacerbates the challenges involved in solving them together. In this paper, we present a novel framework that interleaves task allocation, scheduling, and motion planning. We introduce a search-based approach for trait-based time-extended task allocation named Incremental Task Allocation Graph Search (ITAGS). In contrast to approaches that solve the three problems in sequence, ITAGS’s interleaved approach enables efficient search for allocations while simultaneously satisfying scheduling constraints and accounting for the time taken to execute motion plans. To enable effective interleaving, we develop a convex combination of two search heuristics that optimizes the satisfaction of task requirements as well as the makespan of the associated schedule. We demonstrate the efficacy of ITAGS using detailed ablation studies and comparisons against two state-of-the-art algorithms in a simulated emergency response domain.

Generating robot motion that fulfills multiple tasks simultaneously is challenging due to the geometric constraints imposed on the robot. In this paper, we propose to solve multi-task problems through learning structured policies from human demonstrations. Our structured policy is inspired by RMPflow, a framework for combining subtask policies on different spaces. The policy structure provides the user an interface to 1) specifying the spaces that are directly relevant to the completion of the tasks, and 2) designing policies for certain tasks that do not need to be learned. We derive an end-to-end learning objective that is suitable for the multi-task problem, emphasizing the distance between generated motions and demonstrations measured on task spaces. Furthermore, the motion generated from the learned policy class is guaranteed to be stable. We validate the effectiveness of our proposed learning framework through qualitative and quantitative evaluations on three robotic tasks on a 7-DOF Rethink Sawyer robot.

Multi-robot task allocation (MRTA) problems involve optimizing the allocation of robots to tasks. MRTA problems are known to be challenging when tasks require multiple robots and the team is composed of heterogeneous robots. These challenges are further exacerbated when we need to account for uncertainties encountered in the real-world. In this work, we address coalition formation in heterogeneous multi-robot teams with uncertain capabilities. We specifically focus on tasks that require coalitions to collectively satisfy certain minimum requirements. Existing approaches to uncertainty-aware task allocation either maximize expected pay-off (risk-neutral approaches) or improve worst-case or near-worst-case outcomes (risk-averse approaches). Within the context of our problem, we demonstrate the inherent limitations of unilaterally ignoring or avoiding risk and show that these approaches can in fact reduce the probability of satisfying task requirements. Inspired by models that explain foraging behaviors in animals, we develop a risk-adaptive approach to task allocation. Our approach adaptively switches between risk-averse and risk-seeking behavior in order to maximize the probability of satisfying task requirements. Comprehensive numerical experiments conclusively demonstrate that our risk-adaptive approach outperforms risk-neutral and risk-averse approaches. We also demonstrate the effectiveness of our approach using a simulated multi-robot emergency response scenario.

Recent work has shown results on learning navigation policies for idealized cylinder agents in simulation and transferring them to real wheeled robots. Deploying such navigation policies on legged robots can be challenging due to their complex dynamics, and the large dynamical difference between cylinder agents and legged systems. In this work, we learn hierarchical navigation policies that account for the low-level dynamics of legged robots, such as maximum speed, slipping, contacts, and learn to successfully navigate cluttered indoor environments. To enable transfer of policies learned in simulation to new legged robots and hardware, we learn dynamics-aware navigation policies across multiple robots with robot-specific embeddings. The learned embedding is optimized on new robots, while the rest of the policy is kept fixed, allowing for quick adaptation. We train our policies across three legged robots in simulation - 2 quadrupeds (A1, AlienGo) and a hexapod (Daisy). At test time, we study the performance of our learned policy on two new legged robots in simulation (Laikago, 4-legged Daisy), and one real-world quadrupedal robot (A1). Our experiments show that our learned policy can sample-efficiently generalize to previously unseen robots, and enable sim-to-real transfer of navigation policies for legged robots.

Coordinating human-robot teams requires careful planning and allocation of tasks to the most appropriate agents. This challenge is exacerbated by the fact that, unlike their robot teammates, humans exhibit significant variation in their abilities. Existing work largely ignores this variation in favor of simpler aggregate models, failing to leverage specialized capabilities of different individuals. In this work, we introduce simple cognitive tests for measuring inherent variations in human capabilities related to human-robot teaming, specifically, the ability to maintain situational awareness and to mentally model latent network structures. We then demonstrate that user study participant performance on these cognitive tests is correlated with, and thus is a predictor for, their performance on human-robot teaming tasks. These findings have the potential to improve human-robot teaming algorithms (e.g., task allocation) by providing a mechanism to better leverage individual differences in human agents.

The objective of the knowledge base completion problem is to infer missing information from existing facts in a knowledge base. Prior work has demonstrated the effectiveness of path-ranking based methods, which solve the problem by discovering observable patterns in knowledge graphs, consisting of nodes representing entities and edges representing relations. However, these patterns either lack accuracy because they rely solely on relations or cannot easily generalize due to the direct use of specific entity information. We introduce Attentive Path Ranking, a novel path pattern representation that leverages type hierarchies of entities to both avoid ambiguity and maintain generalization. Then, we present an end-to-end trained attention-based RNN model to discover the new path patterns from data. Experiments conducted on benchmark knowledge base completion datasets WN18RR and FB15k-237 demonstrate that the proposed model outperforms existing methods on the fact prediction task by statistically significant margins of 26% and 10%, respectively. Furthermore, quantitative and qualitative analyses show that the path patterns balance between generalization and discrimination.

As robot autonomy improves, robots are increasingly being considered in the role of autonomous observation systems - free-flying cameras capable of actively tracking human activity within some predefined area of interest. In this work, we formulate the autonomous observation problem through multi-objective optimization, presenting a novel Semi-MDP formulation of the autonomous human observation problem that maximizes observation rewards while accounting for both human- and robot-centric costs. We demonstrate that the problem can be solved with both scalarization-based Multi-Objective MDP methods and Constrained MDP methods, and discuss the relative benefits of each approach. We validate our work on activity tracking using a NASA Astrobee robot operating within a simulated International Space Station environment.

Semantic grasping is the problem of selecting stable grasps that are functionally suitable for specific object manipulation tasks. In order for robots to effectively perform object manipulation, a broad sense of contexts, including object and task constraints, needs to be accounted for. We introduce the Context-Aware Grasping Engine, which combines a novel semantic representation of grasp contexts with a neural network structure based on the Wide & Deep model, capable of capturing complex reasoning patterns. We quantitatively validate our approach against three prior methods on a novel dataset consisting of 14,000 semantic grasps for 44 objects, 7 tasks, and 6 different object states. Our approach outperformed all baselines by statistically significant margins, producing new insights into the importance of balancing memorization and generalization of contexts for semantic grasping. We further demonstrate the effectiveness of our approach on robot experiments in which the presented model successfully achieved 31 of 32 suitable grasps. The code and data are available at: https://github.com/wliu88/railsemanticgrasping.

We contribute a study benchmarking the performance of multiple motion-based learning from demonstration approaches. Given the number and diversity of existing methods, it is critical that comprehensive empirical studies be performed comparing the relative strengths of these techniques. In particular, we evaluate four approaches based on properties an end user may desire for real-world tasks. To perform this evaluation, we collected data from nine participants, across four manipulation tasks. The resulting demonstrations were used to train 180 task models and evaluated on 720 task reproductions on a physical robot. Our results detail how i) complexity of the task, ii) the expertise of the human demonstrator, and iii) the starting configuration of the robot affect task performance. The collected dataset of demonstrations, robot executions, and evaluations are publicly available. Research insights and guidelines are also provided to guide future research and deployment choices about these approaches.

Material recognition can help inform robots about how to properly interact with and manipulate real-world objects. In this paper, we present a multimodal sensing technique, leveraging near-infrared spectroscopy and close-range high resolution texture imaging, that enables robots to estimate the materials of household objects. We release a dataset of high resolution texture images and spectral measurements collected from a mobile manipulator that interacted with 144 house-hold objects. We then present a neural network architecture that learns a compact multimodal representation of spectral measurements and texture images. When generalizing material classification to new objects, we show that this multimodal representation enables a robot to recognize materials with greater performance as compared to prior state-of-the-art approaches. Finally, we present how a robot can combine this high resolution local sensing with images from the robot's head-mounted camera to achieve accurate material classification over a scene of objects on a table.

We address the problem of adapting robot trajectories to improve safety, comfort, and efficiency in humanrobot collaborative tasks. To this end, we propose CoMOTO, a trajectory optimization framework that utilizes stochastic motion prediction to anticipate the human's motion and adapt the robot's joint trajectory accordingly. We design a multiobjective cost function that simultaneously optimizes for i) separation distance, ii) visibility of the end-effector, iii) legibility, iv) efficiency, and v) smoothness. We evaluate CoMOTO against three existing methods for robot trajectory generation when in close proximity to humans. Our experimental results indicate that our approach consistently outperforms existing methods over a combined set of safety, comfort, and efficiency metrics.

To realize effective heterogeneous multi-agent teams, we must be able to leverage individual agents' relative strengths. Recent work has addressed this challenge by introducing trait-based task assignment approaches that exploit the agents' relative advantages. These approaches, however, assume that the agents' traits remain static. Indeed, in real-world scenarios, traits are likely to vary as agents execute tasks. In this paper, we present a transformation-based modeling framework to bridge the gap between state-of-the-art task assignment algorithms and the reality of dynamic traits. We define a transformation as a function that approximates dynamic traits with static traits based on a specific statistical measure. We define different candidate transformations, investigate their effects on different dynamic trait models, and the resulting task performance. Further, we propose a variance-based transformation as a general solution that approximates a variety of dynamic models, eliminating the need for hand specification. Finally, we demonstrate the benefits of reasoning about dynamic traits both in simulation and in a physical experiment involving the game of capture-the-flag.

Machine learning (ML) systems across many application areas are increasingly demonstrating performance that is beyond that of humans. In response to the proliferation of such models, the field of Explainable AI (XAI) has sought to develop techniques that enhance the transparency and interpretability of machine learning methods. In this work, we consider a question not previously explored within the XAI and ML communities: Given a computational system whose performance exceeds that of its human user, can explainable AI capabilities be leveraged to improve the performance of the human? We study this question in the context of the game of Chess, for which computational game engines that surpass the performance of the average player are widely available. We introduce the Rationale-Generating Algorithm, an automated technique for generating rationales for utility-based computational methods, which we evaluate with a multi-day user study against two baselines. The results show that our approach produces rationales that lead to statistically significant improvement in human task performance, demonstrating that rationales automatically generated from an AI's internal task model can be used not only to explain what the system is doing, but also to instruct the user and ultimately improve their task performance.

Robots operating without close human supervision might need to rely on a remote call center of operators for assistance in the event of a failure. In this work, we investigate the effects of providing decision support through diagnosis suggestions, as feedback, and action recommendations, as feedforward, to the human operators. We conduct a 10-condition user study involving 200 participants on Amazon Mechanical Turk to evaluate the effects of providing noisy and noise-free diagnosis suggestions and/or action recommendations to operators. We find that although action recommendations (feedforward) have a greater effect on successful error resolution than diagnosis information (feedback), the feedback likely helps ameliorate the deleterious effects of noise. Therefore, we find that error recovery interfaces should display both diagnosis and action recommendations for maximum effectiveness.

Large teams of heterogeneous agents have the potential to solve complex multi-task problems that are intractable for a single agent working independently. However, solving complex multi-task problems requires leveraging the relative strengths of the different kinds of agents in the team. We present Stochastic TRAit-based Task Assignment (STRATA), a unified framework that models large teams of heterogeneous agents and performs effective task assignments. Specifically, given information on which traits (capabilities) are required for various tasks, STRATA computes the assignments of agents to tasks such that the trait requirements are achieved. Inspired by prior work in robot swarms and biodiversity, we categorize agents into different species (groups) based on their traits. We model each trait as a continuous variable and differentiate between traits that can and cannot be aggregated from different agents. STRATA is capable of reasoning about both species-level and agent-level variability in traits. Further, we define measures of diversity for any given team based on the team’s continuous-space trait model. We illustrate the necessity and effectiveness of STRATA using detailed experiments based in simulation and in a capture-the-flag game environment.

We address the challenge of how a robot can adapt its actions to successfully manipulate objects it has not previously encountered. We introduce Real-time Multisensory Affordance-based Control (RMAC), which enables a robot to adapt existing affordance models using multisensory inputs. We show that using the combination of haptic, audio, and visual information with RMAC allows the robot to learn afforance models and adaptively manipulate two very different objects (drawer, lamp), in multiple novel configurations. Offline evaluations and real-time online evaluations show that RMAC allows the robot to accurately open different drawer configurations and turn-on novel lamps with an average accuracy of 75%.

Autonomous service robots require computational frameworks that allow them to generalize knowledge to new situations in a manner that models uncertainty while scaling to real-world problem sizes. The Robot Common Sense Embedding (RoboCSE) showcases a class of computational frameworks, multi-relational embeddings, that have not been leveraged in robotics to model semantic knowledge. We validate RoboCSE on a realistic home environment simulator (AI2Thor) to measure how well it generalizes learned knowledge about object affordances, locations, and materials. Our experiments show that RoboCSE can perform prediction better than a baseline that uses pre-trained embeddings, such as Word2Vec, achieving statistically significant improvements while using orders of magnitude less memory than our Bayesian Logic Network baseline. In addition, we show that predictions made by RoboCSE are robust to significant reductions in data available for training as well as domain transfer to MatterPort3D, achieving statistically significant improvements over a baseline that memorizes training data.

MacGyvering is defined as creating or repairing something in an inventive or improvised way by utilizing objects that are available at hand. In this paper, we explore a subset of Macgyvering problems involving tool construction, i.e., creating tools from parts available in the environment. We formalize the overall problem domain of tool Macgyvering, introducing three levels of complexity for tool construction and substitution problems, and presenting a novel computational framework aimed at solving one level of the tool Macgyvering problem, specifically contributing a novel algorithm for tool construction based on geometric reasoning. We validate our approach by constructing three tools using a 7-DOF robot arm.

We propose a learning framework, named Multi-Coordinate Cost Balancing (MCCB), to address the problem of acquiring point-to-point movement skills from demonstrations. MCCB encodes demonstrations simultaneously in multiple differential coordinates that specify local geometric properties. MCCB generates reproductions by solving a convex optimization problem with a multi-coordinate cost function and linear constraints on the reproductions, such as initial, target, and via points. Further, since the relative importance of each coordinate system in the cost function might be unknown for a given skill, MCCB learns optimal weighting factors that balance the cost function. We demonstrate the effectiveness of MCCB via detailed experiments conducted on one handwriting dataset and three complex skill datasets.

Active learning agents typically employ a query selection algorithm which solely considers the agent's learning objectives. However, this may be insufficient in more realistic human domains. This work uses imitation learning to enable an agent in a constrained environment to concurrently reason about both its internal learning goals and environmental constraints externally imposed, all within its objective function. Experiments are conducted on a concept learning task to test generalization of the proposed algorithm to different environmental conditions and analyze how time and resource constraints impact efficacy of solving the learning problem. Our findings show the environmentally-aware learning agent is able to statistically outperform all other active learners explored under most of the constrained conditions. A key implication is adaptation for active learning agents to more realistic human environments, where constraints are often externally imposed on the learner.

The proliferation of commercially available social robots is undeniable. As humans and robots interact more closely and frequently, it brings to light issues surrounding how the human feels and perceives when dealing with robots how much do they like the way the interaction occurs, how well do they understand what the robot is trying to communicate, and how comfortable do they feel? Much of this is intertwined within the communication level that the robot uses. In this paper, we evaluate the effect of different robot communication levels - voice only, sound only, or voice and sound - when applied to robots designed with differing levels of anthropomorphism and different purposes, evaluating the resulting impact on human-robot interaction with respect to likeability, understandability and comfort. We evaluate these factors on 13 commercially designed robots. Our results show that in almost all cases, survey responders showed a preference for robots that incorporate more spoken interaction than currently deployed systems.

Collaborative robots that work alongside humans will experience service breakdowns and make mistakes. These robotic failures can cause a degradation of trust between the robot and the community being served. A loss of trust may impact whether a user continues to rely on the robot for assistance. In order to improve the teaming capabilities between humans and robots, forms of communication that aid in developing and maintaining trust need to be investigated. In our study, we identify four forms of communication which dictate the timing of information given and type of initiation used by a robot. We investigate the effect that these forms of communication have on trust with and without robot mistakes during a cooperative task. Participants played a memory task game with the help of a humanoid robot that was designed to make mistakes after a certain amount of time passed. The results showed that participants' trust in the robot was better preserved when that robot offered advice only upon request as opposed to when the robot took initiative to give advice.

The state features available to a robot define the variables on which the learning computation depends. However, little prior work considers feature selection in the context of deploying a general-purpose robot able to learn new tasks. In this work, we explore human-driven feature selection in which a robotic agent can identify useful features with the aid of a human user, by extracting information from users about which features are most informative for discriminating between classes of objects needed for a given task (e.g. sorting groceries). The research questions examine (a) whether a domain expert is able to identify a subset of informative task features, (b) whether human selected features will enable the agent to classify unseen examples as accurately as using computational feature selection, and (c) if the interaction strategy used to elicit the information from the user impacts the quality of resulting feature selection. Toward that end, we conducted a user study with 30 participants on campus, given a multi-class classification task and one of five different approaches for conveying information about informative features to a robot learner. Our findings show that when features are semantically interpretable, human feature selection is effective in LfD scenarios because it is able to outperform computational methods when there is limited training data, yet still remains on-par with computational methods as the training sample size increases.

Active learning literature has explored the selection of optimal queries by a learning agent with respect to given criteria, but prior work in classification has focused only on obtaining labels for queried samples. In contrast, proficient learners, like humans, integrate multiple forms of information during learning. This work seeks to enable an active learner to reason about multiple query types concurrently, aimed at soliciting both instance and feature information from the teacher, and to autonomously arbitrate between queries of different types. We contribute the design of rule-based and decision-theoretic arbitration strategies and evaluate all against baselines of more traditional passive and active learning. Our findings show that all arbitration strategies lead to more efficient learning, compared to the baselines. Moreover, given a dynamically changing environment and constrained questioning budget (typical in human settings), the decision-theoretic strategy statistically outperforms all other methods since it reasons about both what query to make and when to make a query, in order to most effectively utilize its questioning budget.

Learning from Demonstration (LfD) is a popular approach to endowing robots with skills without having to program them by hand. Typically, LfD relies on human demonstrations in clutter-free environments. This prevents the demonstrations from being affected by irrelevant objects, whose influence can obfuscate the true intention of the human or the constraints of the desired skill. However, it is unrealistic to assume that the robot's environment can always be restructured to remove clutter when capturing human demonstrations. To contend with this problem, we develop an importance weighted batch and incremental skill learning approach, building on a recent inference-based technique for skill representation and reproduction. Our approach reduces unwanted environmental influences on the learned skill, while still capturing the salient human behavior. We provide both batch and incremental versions of our approach and validate our algorithms on a 7-DOF JACO2 manipulator with reaching and placing skills.

As robots become increasingly prevalent in human environments, there will inevitably be times when the robot needs to interrupt a human to initiate an interaction. Our work introduces the first interruptibility-aware mobile-robot system, which uses social and contextual cues online to accurately determine when to interrupt a person. We evaluate multiple non-temporal and temporal models on the interruptibility classification task, and show that a variant of Conditional Random Fields (CRFs), the Latent-Dynamic CRF, is the most robust, accurate, and appropriate model for use on our system. Additionally, we evaluate different classification features and show that the observed demeanor of a person can help in interruptibility classification; but in the presence of detection noise, robust detection of object labels as a visual cue to the interruption context can improve interruptibility estimates. Finally, we deploy our system in a large-scale user study to understand the effects of interruptibility-awareness on human-task performance, robot-task performance, and on human interpretation of the robot’s social aptitude. Our results show that while participants are able to maintain task performance, even in the presence of interruptions, interruptibility-awareness improves the robot’s task performance and improves participant social perceptions of the robot.

Robust remote teleoperation of high-DOF manipulators is of critical importance across a wide range of robotics applications. Contemporary robot manipulation interfaces primarily utilize a free-positioning pose specification approach to independently control each axis of translation and orientation in free space. In this work, we present two novel interfaces, constrained positioning and point-and-click, which incorporate scene information, including points-of-interest and local surface geometry, into the grasp specification process. We also present results of a user study evaluation comparing the effects of increased use of scene information in grasp pose specification algorithms for general object manipulation. The results of our study show that constrained positioning and point-and-click significantly outperform the widely used free positioning approach by significantly reducing the number of grasping errors and the number of user interactions required to specify poses. Furthermore, the point-and-click interface significantly increased the number of tasks users were able to complete.

We present a novel solution to the object search problem for domains in which object permanence cannot be assumed and other agents may move objects between locations without the robot's knowledge. We formalize object search as a failure analysis problem and contribute temporal persistence modeling (TPM), an algorithm for probabilistic prediction of the time that an object is expected to remain at a given location given sparse prior observations. We show that probabilistic exponential distributions augmented with a Gaussian component can accurately represent probable object locations and search suggestions based entirely on sparsely made visual observations. We evaluate our work in two domains, a large scale GPS location data set for person tracking, and multi-object tracking on a mobile robot operating in a small-scale household environment over a 2-week period. TPM performance exceeds four baseline methods across all study conditions.

The assumption that we can preprogram robots with all the information necessary for their function becomes impractical as the range of robotics applications grows. One widely proposed solution is the specification of reusable task plans, or recipes, consisting of a sequence of primitive actions. However, the primitive actions, such as unscrewing a nut in a car maintenance task, cannot always be predefined and thus must be learned for a particular robot platform and workspace. In this work, we present a novel algorithm that enables a robot to identify a reusable motion trajectory associated with each primitive action of a plan. Our approach segments the motion data captured during the demonstration of a human performing the given task, while also leveraging the human's verbal cues. We evaluated our algorithms in a pilot study with 6 users executing 90 primitive actions.

When a robot is deployed to a new setting, it must reason about how to accomplish the goals of domain-appropriate tasks within the environment it is situated. We investigate the problem of enabling robots to interactively learn how to perform known tasks in new environments. Each task is composed of a sequence of parameterized actions, which we assume are given to the robot in the form of a task recipe. In order to learn how to ground the task in a new environment, our learner builds classifiers to model each of the parameters (i.e. all unique objects and semantic locations) associated with the task. In evaluation for two tasks across three different environments, our results show that these groundings are both (1) capable of being learned efficiently from demonstrations, and (2) necessary to learn for each new environment.

We present a novel algorithm to identify reusable motion trajectories corresponding to the primitive actions in a human demonstration of a symbolic plan with accompanying narration. Our approach involves a multi-step process starting with time-series pattern mining applied to raw motion-capture data. We evaluated our algorithm on recordings of human motions and showed that it identifies reusable trajectories with 86% of the accuracy of human experts.

Analogies are a fundamental human reasoning pattern that relies on relational similarity. Understanding how analogies are formed facilitates the transfer of knowledge between contexts. The approach presented in this work focuses on obtaining precise interpretations of analogies. We leverage noisy semantic networks to answer and explain a wide spectrum of analogy questions. The core of our contribution, the Semantic Similarity Engine, consists of methods for extracting and comparing graph-contexts that reveal the relational parallelism that analogies are based on, while mitigating uncertainty in the semantic network.We demonstrate these methods in two tasks: answering multiple choice analogy questions and generating human readable analogy explanations. We evaluate our approach on two datasets totaling 600 analogy questions. Our results show reliable performance and low false-positive rate in question answering; human evaluators agreed with 96% of our analogy explanations.

We have developed learning and interaction algorithms to support a human teaching hierarchical task models to a robot using a single demonstration in the context of a mixed-initiative interaction with bi-directional communication. In particular, we have identified and implemented two important heuristics for suggesting task groupings based on the physical structure of the manipulated artifact and on the data flow between tasks. We have evaluated our algorithms with users in a simulated environment and shown both that the overall approach is usable and that the grouping suggestions significantly improve the learning and interaction.

In order for robots to be useful in real world learning scenarios, non-expert human teachers must be able to interact with and teach robots in an intuitive manner. One essential robot capability is wide-area (mobile or nonstationary) pick-and-place tasks. Even in its simplest form, pick-and-place is a hard problem due to uncertainty arising from noisy input demonstrations and non-deterministic real world environments. This work introduces a novel method for goal-based learning from demonstration where we learn over a large corpus of human demonstrated ground truths of placement locations in an unsupervised manner via Gaussian Mixture Models. The goal is to provide a multi-hypothesis solution for a given task description which can later be utilized in the execution of the task itself. In addition to learning the actual arrangements of the items in question, we also autonomously extract which frames of reference are important in each demonstration. We further verify these findings in a subsequent evaluation and execution via a mobile manipulator.

Since its official introduction in 2012, the Robot Web Tools project has grown tremendously as an open-source community, enabling new levels of interoperability and portability across heterogeneous robot systems, devices, and front-end user interfaces. At the heart of Robot Web Tools is the rosbridge protocol as a general means for messaging ROS topics in a client-server paradigm suitable for wide area networks, and human-robot interaction at a global scale through modern web browsers. Building from rosbridge, this paper describes our efforts with Robot Web Tools to advance: 1) human-robot interaction through usable client and visualization libraries for more efficient development of front-end human-robot interfaces, and 2) cloud robotics through more efficient methods of transporting high-bandwidth topics (e.g., kinematic transforms, image streams, and point clouds). We further discuss the significant impact of Robot Web Tools through a diverse set of use cases that showcase the importance of a generic messaging protocol and front-end development systems for human-robot interaction.

Intelligent behavior in robots is implemented through algorithms. Historically, much of algorithmic robotics research strives to compute outputs that achieve mathematically rigid conditions, such as minimizing path length. But today's robots are increasingly being used to empower the daily lives of people, and experience shows that traditional algorithmic approaches are poorly suited for the unpredictable, idiosyncratic, and adaptive nature of human-robot interaction. This raises a need for entirely new computational, mathematical, and technical approaches for robots to better understand and react to humans. The human-friendly robots of the future will need new algorithms, informed from the ground up by HRI research, to generate interpretable, ethical, socially-acceptable behavior, ensure safety around humans, and execute tasks of value to society.

Current approaches to learning partial ordering constraints by demonstration require demonstrating all (or almost all) possible completion orders. We have developed an algorithm that, for plans involving relative placement of objects, learns the partial ordering constraints from a single demonstration by letting the user specify naturally conceived reference frame information. This work is an example of a broader research agenda that involves applying principles of human collaboration to robot learning from demonstration.

Supervision and teleoperation of high degree-of-freedom robots is a complex task due to environmental constraints such as obstacles and limited communication, as well as task specific requirements such as using more than one end-effector at the same time. In this work we present a supervision and teleoperation framework that allows an operator to see the surroundings of a robot in 3D, make necessary adjustments for a dual or single arm manipulation task, preview the task in simulation before execution, and finally execute the task on a real robot. The framework has been applied to the valve turning task of the DARPA Robotics Challenge on the PR2, Hubo2+, and DRCHubo robots.

Object recognition and manipulation are critical in enabling robots to interact with objects in a household environment. Construction of 3D object recognition databases is time and resource intensive, often requiring specialized equipment, and is therefore difficult to apply to robots in the field. We present a system for constructing object models for 3D object recognition and manipulation made possible by advances in web robotics. The database consists of point clouds generated using a novel iterative point cloud registration algorithm, which includes the potential to encode manipulation data and usability characteristics. We validate the system with a crowdsourcing user study and object recognition system designed to work with our object recognition database.

Intelligent object manipulation is critical for a robot to effectively operate in a household environment. There are many grasp planners that can estimate grasps based on object shape, but these approaches often perform poorly because they miss key information about non-visual object characteristics. Object model databases can account for this information, but existing methods for database construction are time and resource intensive. We present an easy-to-use system for constructing a grasp database from crowdsourced demonstrations. The method requires no additional equipment other than the robot itself, and non-expert users can demonstrate grasps through an intuitive web interface, with virtually no training required. We show that the crowdsourced grasps can prove sufficient for object manipulation, and furthermore the demonstration approach outperforms purely vision-based grasp planning approaches for a wide variety of object classes.

In this paper, we report lessons learned through the design of a framework for teleoperating a humanoid robot to perform a manipulation task. We present a software framework for cooperative traded control that enables a team of operators to control a remote humanoid robot over an unreliable communications link. The framework produces statically-stable motion trajectories that are collision-free and respect end-effector pose constraints. After operator confirmation, these trajectories are sent over the data link for execution on the robot. Additionally, we have defined a clear operational procedure for the operators to manage the teleoperation task. We applied our system to the valve turning task in the DARPA Robotics Challenge (DRC). Our framework is able to perform reliably and is resilient to unreliable network conditions, as we demonstrate in a set of test runs performed remotely over the internet. We analyze our approach and discuss lessons learned which may be useful for others when designing such a system.

simultaneously licensed under a Creative Commons Attribution License that allows others to share the work with an acknowledgement of the work's authorship and initial publication in this journal.

CloudPrimer is a tablet-based interactive reading primer that aims to foster early literacy skills and shared parent-child reading through user-targeted discussion topic suggestions. The tablet application records discussions between parents and children as they read a story and leverages this information, in combination with a common sense knowledge base, to develop discussion topic models. The long-term goal of the project is to use such models to provide context-sensitive discussion topic suggestions to parents during the shared reading activity in order to enhance the interactive experience and foster parental engagement in literacy education. In this paper, we present a novel approach for using commonsense reasoning to effectively model topics of discussion in unstructured dialog. We introduce a metric for localizing concepts that the users are interested in at a given moment in the dialog and extract a time sequence of words of interest. We then present algorithms for topic modeling and refinement that leverage semantic knowledge acquired from ConceptNet, a commonsense knowledge base. We evaluate the performance of our algorithms using transcriptions of audio recordings of parent-child pairs interacting with a tablet application, and compare the output of our algorithms to human-generated topics. Our results show that words of interest and discussion topics selected by our algorithm closely match those identified by human readers.

Supporting a wide variety of interaction styles across a diverse set of people is a significant challenge in human-robot interaction (HRI). In this work, we explore a data-driven approach that relies on crowdsourcing as a rich source of interactions that cover a wide repertoire of human behavior. We first develop an online game that requires two players to collaborate to solve a task. One player takes the role of a robot avatar and the other a human avatar, each with a different set of capabilities that must be coordinated to overcome challenges and complete the task. Leveraging the interaction data recorded in the online game, we present a novel technique for data-driven behavior generation using case-based planning for a real robot. We compare the resulting autonomous robot behavior against a Wizard of Oz base case condition in a real-world reproduction of the online game that was conducted at the Boston Museum of Science. Results of a post-study survey of participants indicate that the autonomous robot behavior matched the performance of the human-operated robot in several important measures. We examined video recordings of the real-world game to draw additional insights as to how the novice participants attempted to interact with the robot in a loosely structured collaborative task. We discovered that many of the collaborative interactions were generated in the moment and were driven by interpersonal dynamics, not necessarily by the task design. We explored using bids analysis as a meaningful construct to tap into affective qualities of HRI. An important lesson from this work is that in loosely structured collaborative tasks, robots need to be skillful in handling these in-the-moment interpersonal dynamics, as these dynamics have an important impact on the affective quality of the interaction for people. How such interactions dovetail with more task-oriented policies is an important area for future work, as we anticipate such interactions becoming commonplace in situations where personal robots perform loosely structured tasks in interaction with people in human living spaces.

Many different robot learning from demonstration methods have been applied and tested in various environments recently. Representation of learned plans, tasks and policies often depends on the technique due to method-specific parameters. An agent that is able to switch between representations can apply its knowledge to different algorithms. This flexibility can be useful for a human teacher when training the agent. In this work we present a process to convert learned policies with two specific methods, Confidence-Based Autonomy (CBA) and Interactive Reinforcement Learning (Int-RL), to each other. Our finding suggests that it is possible for an agent to learn a policy with either CBA or Int-RL method and execute the task with the other with the benefit of previously learned knowledge.

Research on robot learning from demonstration has seen significant growth in recent years, but existing evaluations have focused exclusively on algorithmic performance and not on usability factors, especially with respect to naïve users. Here we present findings from a comparative user study in which we asked non-experts to evaluate three distinctively different robot learning from demonstration algorithms - Behavior Networks, Interactive Reinforcement Learning, and Confidence Based Autonomy. Participants in the study showed a preference for interfaces where they controlled the robot directly (teleoperation and guidance) instead of providing retroactive feedback for past actions (reward and correction). Our results show that the best policy performance in most metrics was achieved using the Confidence Based Autonomy algorithm.

Most human interactions with the environment depend on our ability to navigate freely and to use our hands and arms to manipulate objects. Developing natural means of controlling these abilities in humanoid robots can significantly broaden the usability of such platforms. An ideal interface for humanoid robot teleoperation will be inexpensive, person-independent, require no wearable equipment, and will be easy to use, requiring little or no user training.

The ability for robots to engage in interactive behavior with a broad range of people is critical for future development of social robotic applications. In this paper, we propose the use of online games as a means of generating large-scale data corpora for human-robot interaction research in order to create robust and diverse interaction models. We describe a data collection approach based on a multiplayer game that was used to collect movement, action and dialog data from hundreds of online users. We then study how these records of human-human interaction collected in a virtual world can be used to generate contextually correct social and task-oriented behaviors for a robot collaborating with a human in a similar real-world environment. We evaluate the resulting behavior model using a physical robot in the Boston Museum of Science, and show that the robot successfully performs the collaborative task and that its behavior is strongly influenced by patterns in the crowdsourced dataset.

The Interactive Reinforcement Learning algorithm enables a human user to train a robot by providing rewards in response to past actions and anticipatory guidance to guide the selection of future actions. Past work with software agents has shown that incorporating user guidance into the policy learning process through Interactive Reinforcement Learning significantly improves the policy learning time by reducing the number of states the agent explores. We present the first study of Interactive Reinforcement Learning in real-world robotic systems. We report on four experiments that study the effects that teacher guidance and state space size have on policy learning performance. We discuss modifications made to apply Interactive Reinforcement Learning to a real-world system and show that guidance significantly reduces the learning rate, and that its positive effects increase with state space size.

Robot learning from demonstration focuses on algorithms that enable a robot to learn a policy from demonstrations performed by a teacher, typically a human expert. This paper presents an experimental evaluation of two learning from demonstration algorithms, Interactive Reinforcement Learning and Behavior Networks. We evaluate the performance of these algorithms using a humanoid robot and discuss the relative advantages and drawbacks of these methods with respect to learning time, number of demonstrations, ease of implementation and other metrics. Our results show that Behavior Networks rely on a greater degree of domain knowledge and programmer expertise, requiring very precise definitions for behavior pre- and post-conditions. By contrast Interactive RL requires a relatively simple implementation based only on the robot's sensor data and actions. However, Behavior Networks leverage the pre-coded knowledge to effectively reduce learning time and the required number of human interactions to learn the task.

We demonstrate that structured light-based depth sensing with standard perception algorithms can enable mobile peer-to-peer interaction between humans and robots. We posit that the use of recent emerging devices for depth-based imaging can enable robot perception of non-verbal cues in human movement in the face of lighting and minor terrain variations. Toward this end, we have developed an integrated robotic system capable of person following and responding to verbal and non-verbal commands under varying lighting conditions and uneven terrain. The feasibility of our system for peer-to-peer HRI is demonstrated through two trials in indoor and outdoor environments.

We present Confidence-Based Autonomy (CBA), an interactive algorithm for policy learning from demonstration. The CBA algorithm consists of two components which take advantage of the complimentary abilities of humans and computer agents. The first component, Confident Execution, enables the agent to identify states in which demonstration is required, to request a demonstration from the human teacher and to learn a policy based on the acquired data. The algorithm selects demonstrations based on a measure of action selection confidence, and our results show that using Confident Execution the agent requires fewer demonstrations to learn the policy than when demonstrations are selected by a human teacher. The second algorithmic component, Corrective Demonstration, enables the teacher to correct any mistakes made by the agent through additional demonstrations in order to improve the policy and future task performance. CBA and its individual components are compared and evaluated in a complex simulated driving domain. The complete CBA algorithm results in the best overall learning performance, successfully reproducing the behavior of the teacher while balancing the tradeoff between number of demonstrations and number of incorrect actions during learning.

Effective learning from demonstration techniques enable complex robot behaviors to be taught from a small number of demonstrations. A number of recent works have explored interactive approaches to demonstration, in which both the robot and the teacher are able to select training examples. In this paper, we focus on a demonstration selection algorithm used by the robot to identify informative states for demonstration. Existing automated approaches for demonstration selection typically rely on a single threshold value, which is applied to a measure of action confidence. We highlight the limitations of using a single fixed threshold for a specific subset of algorithms, and contribute a method for automatically setting multiple confidence thresholds designed to target domain states with the greatest uncertainty. We present a comparison of our multi-threshold selection method to confidence-based selection using a single fixed threshold, and to manual data selection by a human teacher. Our results indicate that the automated multi-threshold approach significantly reduces the number of demonstrations required to learn the task.

In their interactions with the world robots inevitably face equivalent action choices, situations in which multiple actions are equivalently applicable. In this paper, we address the problem of equivalent action choices in learning from demonstration, a robot learning approach in which a policy is acquired from human demonstrations of the desired behavior. We note that when faced with a choice of equivalent actions, a human teacher often demonstrates an action arbitrarily and does not make the choice consistently over time. The resulting inconsistently labeled training data poses a problem for classification-based demonstration learning algorithms by violating the common assumption that for any world state there exists a single best action. This problem has been overlooked by previous approaches for demonstration learning. In this paper, we present an algorithm that identifies regions of the state space with conflicting demonstrations and enables the choice between multiple actions to be represented explicitly within the robotpsilas policy. An experimental evaluation of the algorithm in a real-world obstacle avoidance domain shows that reasoning about action choices significantly improves the robotpsilas learning performance.

Legged robots, such as the Sony AIBO, create opportunity to design rich motions to be executed in specific situations. In particular, teams involved in robot soccer RoboCup competitions have developed many different motions for kicking the ball. Designing effective motions and determining their effects is a challenging problem that is traditionally approached through a generate and test methodology. In this paper, we present a method we developed for learning the effects of kicking motions. Our procedure acquires models of the kicks in terms of key values that describe their effects on the ball's trajectory, namely the angle and the distance reached. The successful automated acquisition of the models of different kicks is then followed by the incorporation of these models into the behaviors to select the most promising kick in a given state of the world. Using the robot soccer domain, we demonstrate that a robot that takes into account the learned predicted effects of its actions performs significantly better than its counterpart.

Developing fast gaits for legged robots is a difficult task that requires optimizing parameters in a highly irregular, multidimensional space. In the past, walk optimization for quadruped robots, namely the Sony AIBO robot, was done by handtuning the parameterized gaits. In addition to requiring a lot of time and human expertise, this process produced sub-optimal results. Several recent projects have focused on using machine learning to automate the parameter search. Algorithms utilizing Powell's minimization method and policy gradient reinforcement learning have shown significant improvement over previous walk optimization results. In this paper we present a new algorithm for walk optimization based on an evolutionary approach. Unlike previous methods, our algorithm does not attempt to approximate the gradient of the multidimensional space. This makes it more robust to noise in parameter evaluations and avoids prematurely converging to local optima, a problem encountered by both of the previously suggested algorithms. Our evolutionary algorithm matches the best previous learning method, achieving several different walks of high quality. Furthermore, the best learned walks represent an impressive 20% improvement over our own best hand-tuned walks.