Effectively capturing the joint distribution of all agents in a scene is relevant for predicting the true evolution of the scene and in turn providing more accurate information to the decision processes of autonomous vehicles. While new models have been developed for this purpose in recent years, it remains unclear how to best represent the joint distributions particularly from the perspective of the interactions between agents. Thus far there is no clear consensus on how best to represent interactions between agents; whether they should be learned implicitly from data by neural networks, or explicitly modeled using the spatial and temporal relations that are more grounded in human decision-making. This paper aims to study various means of describing interactions within the same network structure and their effect on the final learned joint distributions. Our findings show that more often than not, simply allowing a network to establish interactive connections between agents based on data has a detrimental effect on performance. Instead, having well defined interactions (such as which agent of an agent pair passes first at an intersection) can often bring about a clear boost in performance.

Robot learning from humans has been proposed and researched for several decades as a means to enable robots to learn new skills or adapt existing ones to new situations. Recent advances in AI, including learning approaches like reinforcement learning and architectures like transformers and foundation models, combined with access to massive datasets, have created attractive opportunities to apply those data-hungry techniques to this problem. We argue that the focus on massive amounts of pre-collected data, and the resulting learning paradigm, where humans demonstrate and robots learn in isolation, is overshadowing a specialized area of work we term Human-Interactive Robot Learning (HIRL). This paradigm, wherein robots and humans interact during the learning process , is at the intersection of multiple fields (AI, robotics, human–computer interaction, design and others) and holds unique promise. Using HIRL, robots can achieve greater sample efficiency (as humans can provide task knowledge through interaction), align with human preferences (as humans can guide the robot behavior toward their expectations), and explore more meaningfully and safely (as humans can utilize domain knowledge to guide learning and prevent catastrophic failures). This can result in robotic systems that can more quickly and easily adapt to new tasks in human environments. The objective of this article is to provide a broad and consistent overview of HIRL research and to guide researchers toward understanding the scope of HIRL, and current open or underexplored challenges related to four themes—namely, human, robot learning, interaction, and broader context. The article includes concrete use cases to illustrate the interaction between these challenges and inspire further research according to broad recommendations and a call for action for the growing HIRL community.

This paper introduces a new imitation learning framework based on energy-based generative models capable of learning complex, physics-dependent, robot motion policies through state-only expert motion trajectories. Our algorithm, called Noise-conditioned Energy-based Annealed Rewards (NEAR), constructs several perturbed versions of the expert's motion data distribution and learns smooth, and well-defined representations of the data distribution's energy function using denoising score matching. We propose to use these learnt energy functions as reward functions to learn imitation policies via reinforcement learning. We also present a strategy to gradually switch between the learnt energy functions, ensuring that the learnt rewards are always well-defined in the manifold of policy-generated samples. We evaluate our algorithm on complex humanoid tasks such as locomotion and martial arts and compare it with state-only adversarial imitation learning algorithms like Adversarial Motion Priors (AMP). Our framework sidesteps the optimisation challenges of adversarial imitation learning techniques and produces results comparable to AMP in several quantitative metrics across multiple imitation settings.

In robot manipulation, Reinforcement Learning (RL) often suffers from low sample efficiency and uncertain convergence, especially in large observation and action spaces. Foundation Models (FMs) offer an alternative, demonstrating promise in zero-shot and few-shot settings. However, they can be unreliable due to limited physical and spatial understanding. We introduce ExploRLLM, a method that combines the strengths of both paradigms. In our approach, FMs improve RL convergence by generating policy code and efficient representations, while a residual RL agent compensates for the FMs' limited physical understanding. We show that Explorllm outperforms both policies derived from FMs and RL baselines in table-top manipulation tasks. Additionally, real-world experiments show that the policies exhibit promising zero-shot sim-to-real transfer. Supplementary material is available at https://explorllm.github.io.

This paper presents an Impedance Primitive-augmented hierarchical reinforcement learning framework for efficient robotic manipulation in sequential contact tasks. We leverage this hierarchical structure to sequentially execute behavior primitives with variable stiffness control capabilities for contact tasks. Our proposed approach relies on three key components: an action space enabling variable stiffness control, an adaptive stiffness controller for dynamic stiffness adjustments during primitive execution, and affordance coupling for efficient exploration while encouraging compliance. Through comprehensive training and evaluation, our framework learns efficient stiffness control capabilities and demonstrates improvements in learning efficiency, compositionality in primitive selection, and success rates compared to the state-of-the-art. The training environments include block lifting, door opening, object pushing, and surface cleaning. Real world evaluations further confirm the framework's sim2real capability. This work lays the foundation for more adaptive and versatile robotic manipulation systems, with potential applications in more complex contact-based tasks.

Achieving controlled jumping behaviour for a quadruped robot is a challenging task, especially when introducing passive compliance in mechanical design. This study addresses this challenge via imitation-based deep reinforcement learning with a progressive training process. To start, we learn the jumping skill by mimicking a coarse jumping example generated by model-based trajectory optimization. Subsequently, we generalize the learned policy to broader situations, including various distances in both forward and lateral directions, and then pursue robust jumping in unknown ground unevenness. In addition, without tuning the reward much, we learn the jumping policy for a quadruped with parallel elasticity. Results show that using the proposed method, i) the robot learns versatile jumps by learning only from a single demonstration, ii) the robot with parallel compliance reduces the landing error by 11.1%, saves energy cost by 15.2% and reduces the peak torque by 15.8%, compared to the rigid robot without parallel elasticity, iii) the robot can perform jumps of variable distances with robustness against ground unevenness (maximal ±4cm height perturbations) using only proprioceptive perception.

Learning from Interactive Demonstrations has revolutionized the way non-expert humans teach robots. It is enough to kinesthetically move the robot around to teach pick-and-place, dressing, or cleaning policies. However, the main challenge is correctly generalizing to novel situations, e.g., different surfaces to clean or different arm postures to dress. This article proposes a novel task parameterization and generalization to transport the original robot policy, i.e., position, velocity, orientation, and stiffness. Unlike the state of the art, only a set of keypoints is tracked during the demonstration and the execution, e.g., a point cloud of the surface to clean. We then propose to fit a nonlinear transformation that would deform the space and then the original policy using the paired source and target point sets. The use of function approximators like Gaussian Processes allows us to generalize, or transport, the policy from every space location while estimating the uncertainty of the resulting policy due to the limited task keypoints and the reduced number of demonstrations. We compare the algorithm's performance with state-of-the-art task parameterization alternatives and analyze the effect of different function approximators. We also validated the algorithm on robot manipulation tasks, i.e., different posture arm dressing, different location product reshelving, and different shape surface cleaning. A video of the experiments can be found here:https://youtu.be/bE6uOnAQBLo.

Controlled execution of dynamic motions in quadrupedal robots, especially those with articulated soft bodies, presents a unique set of challenges that traditional methods struggle to address efficiently. In this study, we tackle these issues by relying on a simple yet effective two-stage learning framework to generate dynamic motions for quadrupedal robots. First, a gradient-free evolution strategy is employed to discover simply represented control policies, eliminating the need for a predefined reference motion. Then, we refine these policies using deep reinforcement learning. Our approach enables the acquisition of complex motions like pronking and back-flipping, effectively from scratch. Additionally, our method simplifies the traditionally labour-intensive task of reward shaping, boosting the efficiency of the learning process. Importantly, our framework proves particularly effective for articulated soft quadrupeds, whose inherent compliance and adaptability make them ideal for dynamic tasks but also introduce unique control challenges.

A central challenge in Learning from Demonstration is to generate representations that are adaptable and can generalize to unseen situations. This work proposes to learn such a representation without using task-specific heuristics within the context of multi-reference frame skill learning by superimposing local skills in the global frame. Local policies are first learned by fitting the relative skills with respect to each frame using Gaussian Processes (GPs). Then, another GP, which determines the relevance of each frame for every time step, is trained in a self-supervised manner from a different batch of demonstrations. The uncertainty quantification capability of GPs is exploited to stabilize the local policies and to train the frame relevance in a fully Bayesian way. We validate the method through a dataset of multi-frame tasks generated in simulation and on real-world experiments with a robotic manipulation pick-and-place re-shelving task.We evaluate the performance of our method with two metrics: how close the generated trajectories get to each of the task goals and the deviation between these trajectories and test expert trajectories. According to both of these metrics, the proposed method consistently outperforms the state-of-the-art baseline, Task-Parameterised Gaussian Mixture Model (TPGMM).

Predicting the future behavior of human road users is an important aspect for the development of risk-aware autonomous vehicles. While many models have been developed towards this end, effectively capturing and predicting the variability inherent to human behavior still remains an open challenge. This paper proposes TrajFlow—a new approach for probabilistic trajectory prediction based on Normalizing Flows. We reformulate the problem of capturing distributions over trajectories into capturing distributions over abstracted trajectory features using an autoencoder, simplifying the learning task of the Normalizing Flows. TrajFlow outperforms state-of-the-art behavior prediction models in capturing full trajectory distributions in two synthetic benchmarks with known true distributions, and is competitive on the naturalistic datasets ETH/UCY, rounD, and nuScenes. Our results demonstrate the effectiveness of TrajFlow in probabilistic prediction of human behavior.

Autonomous vehicles currently suffer from a time-inefficient driving style caused by uncertainty about human behavior, which could be improved by accurate and reliable prediction models enabling more efficient trajectory planning. However, the evaluation of such models is commonly over-simplistic, ignoring the asymmetric importance of prediction errors and the heterogeneity of the datasets used for testing. We examine the potential of recasting interactions between vehicles as gap acceptance scenarios and evaluating models in this structured environment. To that end, we develop a framework aiming to facilitate the evaluation of any model, by any metric, and in any scenario. We then apply this framework to state-of-the-art prediction models, which all show themselves to be unreliable in the most safety-critical situations.

Introducing parallel elasticity in the hardware design endows quadrupedal robots with the ability to perform explosive and efficient motions. However, for this kind of articulated soft quadruped, realizing dynamic jumping with robustness against system uncertainties remains a challenging problem. To achieve this, we propose an impact-aware jumping planning and control approach. Specifically, an offline kino-dynamic-type trajectory optimizer is first formulated to achieve compliant 3-D jumping motions, using a novel actuated spring-loaded inverted pendulum (SLIP) model. Then, an optimization-based online landing strategy, including preimpact leg motion modulation and postimpact landing recovery, is designed. The actuated SLIP model, with the capability of explicitly characterizing parallel elasticity, captures the jumping and landing dynamics, making the problem of motion generation/regulation more tractable. Finally, a hybrid torque control consisting of a feedback tracking loop and a feedforward compensation loop is employed for motion control. Experiments demonstrate the ability to accomplish robust 3-D jumping motions with stable landing and recovery. Besides, our approach can be applied to quadrupedal robots with or without additional parallel compliance.

With the aim of further enabling the exploitation of intentional impacts in robotic manipulation, a control framework is presented that directly tackles the challenges posed by tracking control of robotic manipulators that are tasked to perform nominally simultaneous impacts. This framework is an extension of the reference spreading (RS) control framework, in which overlapping ante- and post-impact references that are consistent with impact dynamics are defined. In this work, such a reference is constructed starting from a teleoperation-based approach. By using the corresponding ante- and post-impact control modes in the scope of a quadratic programming control approach, peaking of the velocity error and control inputs due to impacts is avoided while maintaining high tracking performance. With the inclusion of a novel interim mode, we aim to also avoid input peaks and steps when uncertainty in the environment causes a series of unplanned single impacts to occur rather than the planned simultaneous impact. This work in particular presents for the first time an experimental evaluation of RS control on a robotic setup, showcasing its robustness against uncertainty in the environment compared to three baseline control approaches.

Developing physically assistive robots capable of dressing assistance has the potential to significantly improve the lives of the elderly and disabled population. However, most robotics dressing strategies considered a single robot only, which greatly limited the performance of the dressing assistance. In fact, healthcare professionals perform the task bimanually. Inspired by them, we propose a bimanual cooperative scheme for robotic dressing assistance. In the scheme, an interactive robot joins hands with the human thus supporting/guiding the human in the dressing process, while the dressing robot performs the dressing task. We identify a key feature: elbow angle that affects the dressing action and propose an optimal strategy for the interactive robot using the feature. A dressing coordinate based on the posture of the arm is defined to better encode the dressing policy. We validate the interactive dressing scheme with extensive experiments and also an ablation study. The experiment video is available onhttps://sites.google.com/view/bimanualassitdressing/home

Balancing safety and efficiency when planning in crowded scenarios with uncertain dynamics is challenging where it is imperative to accomplish the robot's mission without incurring any safety violations. Typically, chance constraints are incorporated into the planning problem to provide probabilistic safety guarantees by imposing an upper bound on the collision probability of the planned trajectory. Yet, this results in an overly conservative behavior on the grounds that the gap between the obtained risk and the specified upper limit is not explicitly restricted. To address this issue, we propose a real-time capable approach to quantify the risk associated with planned trajectories obtained from multiple probabilistic planners, running in parallel, with different upper bounds of the acceptable risk level. Based on the evaluated risk, the least conservative plan is selected provided that its associated risk is below a specified threshold. In such a way, the proposed approach provides probabilistic safety guarantees by attaining a closer bound to the specified risk, while being applicable to generic uncertainties of moving obstacles. We demonstrate the efficiency of our proposed approach, by improving the performance of a state-of-the-art probabilistic planner, in simulations and experiments using a mobile robot in an environment shared with humans.

In order to deploy robots that could be adapted by non-expert users, interactive imitation learning (IIL) methods must be flexible regarding the interaction preferences of the teacher and avoid assumptions of perfect teachers (oracles), while considering they make mistakes influenced by diverse human factors. In this work, we propose an IIL method that improves the human-robot interaction for non-expert and imperfect teachers in two directions. First, uncertainty estimation is included to endow the agents with a lack of knowledge awareness (epistemic uncertainty) and demonstration ambiguity awareness (aleatoric uncertainty), such that the robot can request human input when it is deemed more necessary. Second, the proposed method enables the teachers to train with the flexibility of using corrective demonstrations, evaluative reinforcements, and implicit positive feedback. The experimental results show an improvement in learning convergence with respect to other learning methods when the agent learns from highly ambiguous teachers. Additionally, in a user study, it was found that the components of the proposed method improve the teaching experience and the data efficiency of the learning process.

Learning from humans allows nonexperts to program robots with ease, lowering the resources required to build complex robotic solutions. Nevertheless, such data-driven approaches often lack the ability to provide guarantees regarding their learned behaviors, which is critical for avoiding failures and/or accidents. In this work, we focus on reaching/point-to-point motions, where robots must always reach their goal, independently of their initial state. This can be achieved by modeling motions as dynamical systems and ensuring that they are globally asymptotically stable. Hence, we introduce a novel Contrastive Learning loss for training deep neural networks (DNN) that, when used together with an Imitation Learning loss, enforces the aforementioned stability in the learned motions. Differently from previous work, our method does not restrict the structure of its function approximator, enabling its use with arbitrary DNNs and allowing it to learn complex motions with high accuracy. We validate it using datasets and a real robot. In the former case, motions are two- and four-dimensional, modeled as first- and second-order dynamical systems. In the latter, motions are three, four, and six-dimensional, of first and second order, and are used to control a 7-DoF robot manipulator in its end effector space and joint space.

Designing controllers for complex robots such as humanoids is not an easy task. Often, researchers hand-tune controllers, but this is a time-consuming approach that yields a single controller which cannot generalize well to varied tasks. This work presents a method which uses the NSGA-II multi-objective optimization algorithm with various training trajectories to output a diverse Pareto set of well-functioning controller weights and gains. The best of these are shown to also work well on the real Talos robot. The learned Pareto front is then used in a Bayesian optimization (BO) algorithm both as a search space and as a source of prior information in the initial mean estimate. This combined learning approach, leveraging the two optimization methods together, finds a suitable parameter set for a new trajectory within 20 trials and outperforms both BO in the continuous parameter search space and random search along the precomputed Pareto front. The few trials required for this formulation of BO suggest that it could feasibly be applied on the physical robot using a Pareto front generated in simulation.

Existing Deep Learning (DL) frameworks typically do not provide ready-to-use solutions for robotics, where very specific learning, reasoning, and embodiment problems exist. Their relatively steep learning curve and the different methodologies employed by DL compared to traditional approaches, along with the high complexity of DL models, which often leads to the need of employing specialized hardware accelerators, further increase the effort and cost needed to employ DL models in robotics. Also, most of the existing DL methods follow a static inference paradigm, as inherited by the traditional computer vision pipelines, ignoring active perception, which can be employed to actively interact with the environment in order to increase perception accuracy. In this paper, we present the Open Deep Learning Toolkit for Robotics (OpenDR). OpenDR aims at developing an open, non-proprietary, efficient, and modular toolkit that can be easily used by robotics companies and research institutions to efficiently develop and deploy AI and cognition technologies to robotics applications, providing a solid step towards addressing the aforementioned challenges. We also detail the design choices, along with an abstract interface that was created to overcome these challenges. This interface can describe various robotic tasks, spanning beyond traditional DL cognition and inference, as known by existing frameworks, incorporating openness, homogeneity and robotics-oriented perception e.g., through active perception, as its core design principles.

Aleatoric uncertainty estimation, based on the observed training data, is applied for the detection of conflicts in a demonstration data set. The particular focus of this paper is the resolution of conflicting data resulting from scenarios with equivalent action choices, such as obstacle avoidance, path planning or multiple joint configurations. In terms of the estimated uncertainty, the proposed algorithm aims to decrease this otherwise irreducible value through direct alteration of the accrued data set and to provide data that a policy-learning neural network is able to fit appropriately. The proposed algorithm was validated with real robot scenarios while learning from inconsistent demonstrations, where the resulting policies consistently achieved their prescribed objectives. A video showing our method and experiments can be found at: https://youtu.be/oGYnzlW9Ncw.

Teaching robots how to apply forces according to our preferences is still an open challenge that has to be tackled from multiple engineering perspectives. This paper studies how to learn variable impedance policies where both the Cartesian stiffness and the attractor can be learned from human demonstrations and corrections with a user-friendly interface. The presented framework, named ILoSA, uses Gaussian Processes for policy learning, identifying regions of uncertainty and allowing interactive corrections, stiffness modulation and active disturbance rejection. The experimental evaluation of the framework is carried out on a Franka-Emika Panda in four separate cases with unique force interaction properties: 1) pulling a plug wherein a sudden force discontinuity occurs upon successful removal of the plug, 2) pushing a box where a sustained force is required to keep the robot in motion, 3) wiping a whiteboard in which the force is applied perpendicular to the direction of movement, and 4) inserting a plug to verify the usability for precision-critical tasks in an experimental validation performed with non-expert users.

This paper presents DeepKoCo, a novel modelbased agent that learns a latent Koopman representation from images. This representation allows DeepKoCo to plan efficiently using linear control methods, such as linear model predictive control. Compared to traditional agents, DeepKoCo learns taskrelevant dynamics, thanks to the use of a tailored lossy autoencoder network that allows DeepKoCo to learn latent dynamics that reconstruct and predict only observed costs, rather than all observed dynamics. As our results show, DeepKoCo achieves a similar final performance as traditional model-free methods on complex control tasks, while being considerably more robust to distractor dynamics, making the proposed agent more amenable for real-life applications.

Teaching a robot how to navigate in a new environment only from the sensor input in an end-to-end fashion is still an open challenge with much attention from industry and academia. This paper proposes an algorithm with the name “Learning Interactively to Resolve Ambiguity” (LIRA) that tackles the problem of sensor policy fusion extending state- of-the-art methods by employing ambiguity awareness in the decision-making and solving it using active and interactive querying of the human expert. LIRA, in fact, employs Gaussian Processes for the estimation of the policy’s confidence and investigates the ambiguity due to the disagreement between the single sensor policies on the desired action to take. LIRA aims to make the teaching of new policies easier, learning from human demonstrations and correction.The experiments show that LIRA can be used for learning a sensor-fused policy from scratch or also leveraging the knowledge of existing single sensor policies. The experiments focus on the estimation of the human interventions required for teaching a successful navigation policy.

This paper presents a method to incorporate ergonomics into the optimization of action sequences for bi-manual human-robot cooperation tasks with continuous physical interaction. Our first contribution is a novel computational model of the human that allows prediction of an ergonomics assessment corresponding to each step in a task. The model is learned from human motion capture data in order to predict the human pose as realistically as possible. The second contribution is a combination of this prediction model with an informed graph search algorithm, which allows computation of human-robot cooperative plans with improved ergonomics according to the incorporated method for ergonomic assessment. The concepts have been evaluated in simulation and in a small user study in which the subjects manipulate a large object with a 32 DoF bimanual mobile robot as partner. For all subjects, the ergonomic-enhanced planner shows their reduced ergonomic cost compared to a baseline planner.

Deep Reinforcement Learning (DRL) has become a powerful methodology to solve complex decision-making problems. However, DRL has several limitations when used in real-world problems (e.g., robotics applications). For instance, long training times are required and cannot be accelerated in contrast to simulated environments, and reward functions may be hard to specify/model and/or to compute. Moreover, the transfer of policies learned in a simulator to the real-world has limitations (reality gap). On the other hand, machine learning methods that rely on the transfer of human knowledge to an agent have shown to be time efficient for obtaining well performing policies and do not require a reward function. In this context, we analyze the use of human corrective feedback during task execution to learn policies with high-dimensional state spaces, by using the D-COACH framework, and we propose new variants of this framework. D-COACH is a Deep Learning based extension of COACH (COrrective Advice Communicated by Humans), where humans are able to shape policies through corrective advice. The enhanced version of DCOACH, which is proposed in this paper, largely reduces the time and effort of a human for training a policy. Experimental results validate the efficiency of the D-COACH framework in three different problems (simulated and with real robots), and show that its enhanced version reduces the human training effort considerably, and makes it feasible to learn policies within periods of time in which a DRL agent do not reach any improvement.

This paper presents a system for cooperatively manipulating large objects between a human and a robot. This physical interaction system is designed to handle, transport, or manipulate large objects of different shapes in cooperation with a human. Unique points are the bi-manual physical cooperation, the sequential characteristic of the cooperation including contact changes, and a novel architecture combining force interaction cues, interactive search-based planning, and online trajectory and motion generation. The resulting system implements a mixed initiative collaboration strategy, deferring to the human when his intentions are unclear, and driving the task once understood. This results in an easy and intuitive human-robot interaction. It is evaluated in simulations and on a bi-manual mobile robot with 32 degrees of freedom.

Experience replay is a technique that allows off-policy reinforcement-learning methods to reuse past experiences. The stability and speed of convergence of reinforcement learning, as well as the eventual performance of the learned policy, are strongly dependent on the experiences being replayed. Which experiences are replayed depends on two important choices. The first is which and how many experiences to retain in the experience replay buffer. The second choice is how to sample the experiences that are to be replayed from that buffer. We propose new methods for the combined problem of experience retention and experience sampling. We refer to the combination as experience selection. We focus our investigation specifically on the control of physical systems, such as robots, where exploration is costly. To determine which experiences to keep and which to replay, we investigate different proxies for their immediate and long-term utility. These proxies include age, temporal difference error and the strength of the applied exploration noise. Since no currently available method works in all situations, we propose guidelines for using prior knowledge about the characteristics of the control problem at hand to choose the appropriate experience replay strategy.

Recent years have seen a growing interest in the use of deep neural networks as function approximators in reinforcement learning. In this paper, an experience replay method is proposed that ensures that the distribution of the experiences used for training is between that of the policy and a uniform distribution. Through experiments on a magnetic manipulation task it is shown that the method reduces the need for sustained exhaustive exploration during learning. This makes it attractive in scenarios where sustained exploration is in-feasible or undesirable, such as for physical systems like robots and for life long learning. The method is also shown to improve the generalization performance of the trained policy, which can make it attractive for transfer learning. Finally, for small experience databases the method performs favorably when compared to the recently proposed alternative of using the temporal difference error to determine the experience sample distribution, which makes it an attractive option for robots with limited memory capacity.

Learning sequential force interaction tasks from kinesthetic demonstrations is a promising approach to transfer human manipulation abilities to a robot. In this paper we propose a novel concept to decompose such demonstrations into a set of Movement Primitives (MPs). The decomposition is based on a probability distribution we call Directional Normal Distribution (DND). To capture the sequential properties of the manipulation task, we model the demonstrations with a Hidden Markov Model (HMM). Here, we employ mixtures of DNDs as the HMM's output emissions. The combination of HMMs and mixtures of DNDs allows to infer the MP's composition, i.e., its coordinate frames, control variables and target coordinates from the demonstration data. In addition, it permits to determine an appropriate number of MPs that explains the demonstrations best. We evaluate the approach on kinesthetic demonstrations of a light bulb unscrewing task. Decomposing the task leads to intuitive and meaningful MPs that reflect the natural structure of the task.

Kinesthetic teaching is a promising approach to acquire robot skills in an intuitive way. This paper focuses on learning skills that do not solely rely on kinematics but also need to take into account interaction forces. We present three novel concepts towards learning such force interaction skills. Firstly, we determine segments from a small number of continuous kinesthetic demonstrations using contact information. Secondly, we associate each segment with a movement primitive, and determine its composition, i.e., the control variables and reference frames that allow to reproduce the demonstrated task. Lastly, we propose a concept to determine the transitions between the primitives during reproduction. The proposed methods are evaluated on a box pulling and flipping task, and show very good generalization abilities for objects with different geometries, and situations with different object arrangements.

Classical approaches towards learning coordinated movement tasks often represent a movement in a sequential and exclusive fashion. Introducing concurrency allows to decompose such tasks into a number of separate sequences, for instance for two different end-effectors. While this results in a compact and generic representation of the individual movement primitives (MPs), it is a hard problem to learn their temporal and causal organization. This paper presents a concept for learning movement tasks that require the coordination of several controlled effectors of a robot. We firstly introduce a concept to learn and estimate the progress of individual MPs from a low number of demonstrations. Secondly, we propose a representation of the task that incorporates several concurrent sequences of MPs. Combining these two elements allows to learn and reproduce coordinated bi-manual movement tasks robustly. The synchronization of the concurrent MPs is achieved implicitly using the progress prediction. The approach is evaluated in two simulation studies with a 25 degrees of freedom two-arm robot performing a pick-and-place task.

We present an approach for learning sequential robot skills through kinesthetic teaching. The demonstrations are represented by a sequence graph. Finding the transitions between consecutive basic movements is treated as classification problem where both Support Vector Machines and Gaussian Mixture Models are evaluated as classifiers. We show how the observed primitive order of all demonstrations can help to improve the movement reproduction by restricting the classification outcome to the currently executed primitive and its possible successors in the graph. The approach is validated with an experiment in which a 7-DOF Barrett WAM robot learns to unscrew a light bulb.

In this video, we present approaches for learning throwing and catching skills. We first show how a hitting skill (i.e., table tennis) can be learned using a combination of imitation and reinforcement learning. This hitting skill is subsequently generalized to a catching skill. Secondly, we show how a robot can adapt a throwing skill to new targets. Finally, we demonstrate that a BioRob and a Barrett WAM can play catch together using the previously acquired skills.

Abstract—Many complex robot motor skills can be represented using elementary movements, and there exist efficient techniques for learning parametrized motor plans using demonstrations and self-improvement. However, in many cases, the robot currently needs to learn a new elementary movement even if a parametrized motor plan exists that covers a similar, related situation. Clearly, a method is needed that modulates the elementary movement through the meta-parameters of its representation. In this paper, we show how to learn such mappings from circumstances to meta-parameters using reinforcement learning. We introduce an appropriate reinforcement learning algorithm based on a kernelized version of the reward-weighted regression. We compare this algorithm to several previous methods on a toy example and show that it performs well in comparison to standard algorithms. Subsequently, we show two robot applications of the presented setup; i.e., the generalization of throwing movements in darts, and of hitting movements in table tennis. We show that both tasks can be learned successfully using simulated and real robots. I.

Many motor skills consist of many lower level elementary movements that need to be sequenced in order to achieve a task. In order to learn such a task, both the primitive movements as well as the higher-level strategy need to be acquired at the same time. In contrast, most learning approaches focus either on learning to combine a fixed set of options or to learn just single options. In this paper, we discuss a new approach that allows improving the performance of lower level actions while pursuing a higher level task. The presented approach is applicable to learning a wider range motor skills, but in this paper, we employ it for learning games where the player wants to improve his performance at the individual actions of the game while still performing well at the strategy level game. We propose to learn the lower level actions using Cost-regularized Kernel Regression and the higher level actions using a form of Policy Iteration. The two approaches are coupled by their transition probabilities. We evaluate the approach on a side-stall-style throwing game both in simulation and with a real BioRob.

Hitting and batting tasks, such as tennis forehands, ping-pong strokes, or baseball batting, depend on predictions where the ball can be intercepted and how it can properly be returned to the opponent. These predictions get more accurate over time, hence the behaviors need to be continuously modified. As a result, movement templates with a learned global shape need to be adapted during the execution so that the racket reaches a target position and velocity that will return the ball over to the other side of the net or court. It requires altering learned movements to hit a varying target with the necessary velocity at a specific instant in time. Such a task cannot be incorporated straightforwardly in most movement representations suitable for learning. For example, the standard formulation of the dynamical system based motor primitives (introduced by Ijspeert et al. [1]) does not satisfy this property despite their flexibility which has allowed learning tasks ranging from locomotion to kendama. In order to fulfill this requirement, we reformulate the Ijspeert framework to incorporate the possibility of specifying a desired hitting point and a desired hitting velocity while maintaining all advantages of the original formulation. We show that the proposed movement template formulation works well in two scenarios, i.e., for hitting a ball on a string with a table tennis racket at a specified velocity and for returning balls launched by a ball gun successfully over the net using forehand movements. All experiments were carried out on a Barrett WAM using a four camera vision system.

Although human beings see and move slower than table tennis or baseball robots, they manage to outperform such robot systems. One important aspect of this better performance is the human movement generation. In this paper, we study trajectory generation for table tennis from a biomimetic point of view. Our focus lies on generating efficient stroke movements capable of mastering variations in the environmental conditions, such as changing ball speed, spin and position. We study table tennis from a human motor control point of view. To make headway towards this goal, we construct a trajectory generator for a single stroke using the discrete movement stages hypothesis and the virtual hitting point hypothesis to create a model that produces a human-like stroke movement. We verify the functionality of the trajectory generator for a single forehand stroke both in a simulation and using a real Barrett WAM™.

Reinforcement Learning is an essential ability for robots to learn new motor skills. Nevertheless, few methods scale into the domain of anthropomorphic robotics. In order to improve in terms of efficiency, the problem is reduced onto reward-weighted imitation. By doing so, we are able to generate a framework for policy learning which both unifies previous reinforcement learning approaches and allows the derivation of novel algorithms. We show our two most relevant applications both for motor primitive learning (e.g., a complex Ball-in-a-Cup task using a real Barrett WAM robot arm) and learning task-space control.

The acquisition and self-improvement of novel motor skills is among the most important problems in robotics. Motor primitives offer one of the most promising frameworks for the application of machine learning techniques in this context. Employing an improved form of the dynamic systems motor primitives originally introduced by Ijspeert et al. [2], we show how both discrete and rhythmic tasks can be learned using a concerted approach of both imitation and reinforcement learning. For doing so, we present both learning algorithms and representations targeted for the practical application in robotics. Furthermore, we show that it is possible to include a start-up phase in rhythmic primitives. We show that two new motor skills, i.e., Ball-in-a-Cup and Ball-Paddling, can be learned on a real Barrett WAM robot arm at a pace similar to human learning while achieving a significantly more reliable final performance.

Dynamic system-based motor primitives have enabled robots to learn complex tasks ranging from Tennis-swings to locomotion. However, to date there have been only few extensions which have incorporated perceptual coupling to variables of external focus, and, furthermore, these modifications have relied upon handcrafted solutions. Humans learn how to couple their movement primitives with external variables. Clearly, such a solution is needed in robotics. In this paper, we propose an augmented version of the dynamic systems motor primitives which incorporates perceptual coupling to an external variable. The resulting perceptually driven motor primitives include the previous primitives as a special case and can inherit some of their interesting properties. We show that these motor primitives can perform complex tasks such a Ball-in-a-Cup or Kendama task even with large variances in the initial conditions where a skilled human player would be challenged. For doing so, we initialize the motor primitives in the traditional way by imitation learning without perceptual coupling. Subsequently, we improve the motor primitives using a novel reinforcement learning method which is particularly well-suited for motor primitives.

Motor primitives or motion templates have become an important concept for both modeling human motor control as well as generating robot behaviors using imitation learning. Recent impressive results range from humanoid robot movement generation to timing models of human motions. The automatic generation of skill libraries containing multiple motion templates is an important step in robot learning. Such a skill learning system needs to cluster similar movements together and represent each resulting motion template as a generative model which is subsequently used for the execution of the behavior by a robot system. In this paper, we show how human trajectories captured as multidimensional time-series can be clustered using Bayesian mixtures of linear Gaussian state-space models based on the similarity of their dynamics. The appropriate number of templates is automatically determined by enforcing a parsimonious parametrization. As the resulting model is intractable, we introduce a novel approximation method based on variational Bayes, which is especially designed to enable the use of efficient inference algorithms. On recorded human Balero movements, this method is not only capable of finding reasonable motion templates but also yields a generative model which works well in the execution of this complex task on a simulated anthropomorphic SARCOS arm.

Many motor skills in humanoid robotics can be learned using parametrized motor primitives as done in imitation learning. However, most interesting motor learning problems are high-dimensional reinforcement learning problems often beyond the reach of current methods. In this paper, we extend previous work on policy learning from the immediate reward case to episodic reinforcement learning. We show that this results into a general, common framework also connected to policy gradient methods and yielding a novel algorithm for policy learning by assuming a form of exploration that is particularly well-suited for dynamic motor primitives. The resulting algorithm is an EM-inspired algorithm applicable in complex motor learning tasks. We compare this algorithm to alternative parametrized policy search methods and show that it outperforms previous methods. We apply it in the context of motor learning and show that it can learn a complex Ball-in-a-Cup task using a real Barrett WAM robot arm.