Ramya Ramakrishnan, Hashan Narangodage, Mauro Schilman, Kilian Weinberger, Ryan McDonald. Proceedings of the 2022 Conference of the North American Chapter of the Association for Computational Linguistics: Human Language Technologies. 2022.

Agents trained in simulation may make errors when performing actions in the real world due to mismatches between training and execution environments. These mistakes can be dangerous and difficult for the agent to discover because the agent is unable to predict them a priori. In this work, we propose the use of oracle feedback to learn a predictive model of these blind spots in order to reduce costly errors in real-world applications. We focus on blind spots in reinforcement learning (RL) that occur due to incomplete state representation: when the agent lacks necessary features to represent the true state of the world, and thus cannot distinguish between numerous states. We formalize the problem of discovering blind spots in RL as a noisy supervised learning problem with class imbalance. Our system learns models for predicting blind spots within unseen regions of the state space by combining techniques for label aggregation, calibration, and supervised learning. These models take into consideration noise emerging from different forms of oracle feedback, including demonstrations and corrections. We evaluate our approach across two domains and demonstrate that it achieves higher predictive performance than baseline methods, and also that the learned model can be used to selectively query an oracle at execution time to prevent errors. We also empirically analyze the biases of various feedback types and how these biases influence the discovery of blind spots. Further, we include analyses of our approach that incorporate relaxed initial optimality assumptions. (Interestingly, relaxing the assumptions of an optimal oracle and an optimal simulator policy helped our models to perform better.) We also propose extensions to our method that are intended to improve performance when using corrections and demonstrations data.

Simulators are being increasingly used to train agents before deploying them in real-world environments. While training in simulation provides a cost-effective way to learn, poorly modeled aspects of the simulator can lead to costly mistakes, or blind spots. While humans can help guide an agent towards identifying these error regions, humans themselves have blind spots and noise in execution. We study how learning about blind spots of both can be used to manage hand-off decisions when humans and agents jointly act in the real-world in which neither of them are trained or evaluated fully. The formulation assumes that agent blind spots result from representational limitations in the simulation world, which leads the agent to ignore important features that are relevant for acting in the open world. Our approach for blind spot discovery combines experiences collected in simulation with limited human demonstrations. The first step applies imitation learning to demonstration data to identify important features that the human is using but that the agent is missing. The second step uses noisy labels extracted from action mismatches between the agent and the human across simulation and demonstration data to train blind spot models. We show through experiments on two domains that our approach is able to learn a succinct representation that accurately captures blind spot regions and avoids dangerous errors in the real world through transfer of control between the agent and the human.

In this work, we design and evaluate a computational learning model that enables a human-robot team to co-develop joint strategies for performing novel tasks that require coordination. The joint strategies are learned through "perturbation training," a human team-training strategy that requires team members to practice variations of a given task to help their team generalize to new variants of that task. We formally define the problem of human-robot perturbation training and develop and evaluate the first end-to-end framework for such training, which incorporates a multi-agent transfer learning algorithm, human-robot co-learning framework and communication protocol. Our transfer learning algorithm, Adaptive Perturbation Training (AdaPT), is a hybrid of transfer and reinforcement learning techniques that learns quickly and robustly for new task variants. We empirically validate the benefits of AdaPT through comparison to other hybrid reinforcement and transfer learning techniques aimed at transferring knowledge from multiple source tasks to a single target task. We also demonstrate that AdaPT's rapid learning supports live interaction between a person and a robot, during which the human-robot team trains to achieve a high level of performance for new task variants. We augment AdaPT with a co-learning framework and a computational bi-directional communication protocol so that the robot can co-train with a person during live interaction. Results from large-scale human subject experiments (n=48) indicate that AdaPT enables an agent to learn in a manner compatible with a human's own learning process, and that a robot undergoing perturbation training with a human results in a high level of team performance. Finally, we demonstrate that human-robot training using AdaPT in a simulation environment produces effective performance for a team incorporating an embodied robot partner.

We present a framework for automatically learning human user models from joint-action demonstrations that enables a robot to compute a robust policy for a collaborative task with a human. First, the demonstrated action sequences are clustered into different human types using an unsupervised learning algorithm. A reward function is then learned for each type through the employment of an inverse reinforcement learning algorithm. The learned model is then incorporated into a mixed-observability Markov decision process (MOMDP) formulation, wherein the human type is a partially observable variable. With this framework, we can infer online the human type of a new user that was not included in the training set, and can compute a policy for the robot that will be aligned to the preference of this user. In a human subject experiment (n=30), participants agreed more strongly that the robot anticipated their actions when working with a robot incorporating the proposed framework (p<0.01), compared to manually annotating robot actions. In trials where participants faced difficulty annotating the robot actions to complete the task, the proposed framework significantly improved team efficiency (p<0.01). The robot incorporating the framework was also found to be more responsive to human actions compared to policies computed using a hand-coded reward function by a domain expert (p<0.01). These results indicate that learning human user models from joint-action demonstrations and encoding them in a MOMDP formalism can support effective teaming in human-robot collaborative tasks.

In this poster, we summarize the results of a user study that investigates the relative extent to which people are able to maintain spatial awareness when exploring a virtual environment using a motorized wheelchair. We asked 24 participants to travel through a 24' wide, circularly symmetric virtual room, searching the contents of 16 randomly positioned and oriented boxes to locate 8 hidden targets, using each of the following four locomotion methods: real walking; virtual translation with real rotation while standing and using a body-worn joystick; real driving in a motorized wheelchair; and virtual translation with real rotation while sitting in a swivel chair with a joystick mounted on one of its arms. We computed four measures of search efficiency: distance travelled, search time, number of targets revisited, and proportion of trials with no revisits. We found that participants performed significantly better, overall, with real walking than with either of the methods that involved virtual translation, while performance with the wheelchair was intermediate. These results confirm that a mobile, motorized wheelchair interface offers some advantages over stationary joystick travel.