Model-Based Uncertainty in Value Functions

We consider the problem of quantifying uncertainty over expected cumulative rewards in model-based reinforcement learning. In particular, we focus on characterizing the variance over values induced by a distribution over MDPs. Previous work upper bounds the posterior variance over values by solving a so-called uncertainty Bellman equation, but the over-approximation may result in inefficient exploration. We propose a new uncertainty Bellman equation whose solution converges to the true posterior variance over values and explicitly characterizes the gap in previous work. Moreover, our uncertainty quantification technique is easily integrated into common exploration strategies and scales naturally beyond the tabular setting by using standard deep reinforcement learning architectures. Experiments in difficult exploration tasks, both in tabular and continuous control settings, show that our sharper uncertainty estimates improve sample-efficiency.

Diminishing Return of Value Expansion Methods in Model-Based Reinforcement Learning

Model-based reinforcement learning is one approach to increase sample efficiency. However, the accuracy of the dynamics model and the resulting compounding error over modelled trajectories are commonly regarded as key limitations. A natural question to ask is: How much more sample efficiency can be gained by improving the learned dynamics models? Our paper empirically answers this question for the class of model-based value expansion methods in continuous control problems. Value expansion methods should benefit from increased model accuracy by enabling longer rollout horizons and better value function approximations. Our empirical study, which leverages oracle dynamics models to avoid compounding model errors, shows that (1) longer horizons increase sample efficiency, but the gain in improvement decreases with each additional expansion step, and (2) the increased model accuracy only marginally increases the sample efficiency compared to learned models with identical horizons. Therefore, longer horizons and increased model accuracy yield diminishing returns in terms of sample efficiency. These improvements in sample efficiency are particularly disappointing when compared to model-free value expansion methods. Even though they introduce no computational overhead, we find their performance to be on-par with model-based value expansion methods. Therefore, we conclude that the limitation of model-based value expansion methods is not the model accuracy of the learned models. While higher model accuracy is beneficial, our experiments show that even a perfect model will not provide an un-rivalled sample efficiency but that the bottleneck lies elsewhere.

A Human-Centered Safe Robot Reinforcement Learning Framework with Interactive Behaviors

Deployment of reinforcement learning algorithms for robotics applications in the real world requires ensuring the safety of the robot and its environment. Safe robot reinforcement learning (SRRL) is a crucial step towards achieving human-robot coexistence. In this paper, we envision a human-centered SRRL framework consisting of three stages: safe exploration, safety value alignment, and safe collaboration. We examine the research gaps in these areas and propose to leverage interactive behaviors for SRRL. Interactive behaviors enable bi-directional information transfer between humans and robots, such as conversational robot ChatGPT. We argue that interactive behaviors need further attention from the SRRL community. We discuss four open challenges related to the robustness, efficiency, transparency, and adaptability of SRRL with interactive behaviors.

LS-IQ: Implicit Reward Regularization for Inverse Reinforcement Learning

Recent methods for imitation learning directly learn a $Q$-function using an implicit reward formulation rather than an explicit reward function. However, these methods generally require implicit reward regularization to improve stability and often mistreat absorbing states. Previous works show that a squared norm regularization on the implicit reward function is effective, but do not provide a theoretical analysis of the resulting properties of the algorithms. In this work, we show that using this regularizer under a mixture distribution of the policy and the expert provides a particularly illuminating perspective: the original objective can be understood as squared Bellman error minimization, and the corresponding optimization problem minimizes a bounded $\chi^2$-Divergence between the expert and the mixture distribution. This perspective allows us to address instabilities and properly treat absorbing states. We show that our method, Least Squares Inverse Q-Learning (LS-IQ), outperforms state-of-the-art algorithms, particularly in environments with absorbing states. Finally, we propose to use an inverse dynamics model to learn from observations only. Using this approach, we retain performance in settings where no expert actions are available.

Fast Kinodynamic Planning on the Constraint Manifold with Deep Neural Networks

Motion planning is a mature area of research in robotics with many well-established methods based on optimization or sampling the state space, suitable for solving kinematic motion planning. However, when dynamic motions under constraints are needed and computation time is limited, fast kinodynamic planning on the constraint manifold is indispensable. In recent years, learning-based solutions have become alternatives to classical approaches, but they still lack comprehensive handling of complex constraints, such as planning on a lower-dimensional manifold of the task space while considering the robot's dynamics. This paper introduces a novel learning-to-plan framework that exploits the concept of constraint manifold, including dynamics, and neural planning methods. Our approach generates plans satisfying an arbitrary set of constraints and computes them in a short constant time, namely the inference time of a neural network. This allows the robot to plan and replan reactively, making our approach suitable for dynamic environments. We validate our approach on two simulated tasks and in a demanding real-world scenario, where we use a Kuka LBR Iiwa 14 robotic arm to perform the hitting movement in robotic Air Hockey.

Learning-based Design and Control for Quadrupedal Robots with Parallel-Elastic Actuators

Parallel-elastic joints can improve the efficiency and strength of robots by assisting the actuators with additional torques. For these benefits to be realized, a spring needs to be carefully designed. However, designing robots is an iterative and tedious process, often relying on intuition and heuristics. We introduce a design optimization framework that allows us to co-optimize a parallel elastic knee joint and locomotion controller for quadrupedal robots with minimal human intuition. We design a parallel elastic joint and optimize its parameters with respect to the efficiency in a model-free fashion. In the first step, we train a design-conditioned policy using model-free Reinforcement Learning, capable of controlling the quadruped in the predefined range of design parameters. Afterwards, we use Bayesian Optimization to find the best design using the policy. We use this framework to optimize the parallel-elastic spring parameters for the knee of our quadrupedal robot ANYmal together with the optimal controller. We evaluate the optimized design and controller in real-world experiments over various terrains. Our results show that the new system improves the torque-square efficiency of the robot by 33% compared to the baseline and reduces maximum joint torque by 30% without compromising tracking performance. The improved design resulted in 11% longer operation time on flat terrain.

Visual Tactile Sensor Based Force Estimation for Position-Force Teleoperation

Vision-based tactile sensors have gained extensive attention in the robotics community. The sensors are highly expected to be capable of extracting contact information i.e. haptic information during in-hand manipulation. This nature of tactile sensors makes them a perfect match for haptic feedback applications. In this paper, we propose a contact force estimation method using the vision-based tactile sensor DIGIT, and apply it to a position-force teleoperation architecture for force feedback. The force estimation is done by building a depth map for DIGIT gel surface deformation measurement and applying a regression algorithm on estimated depth data and ground truth force data to get the depth-force relationship. The experiment is performed by constructing a grasping force feedback system with a haptic device as a leader robot and a parallel robot gripper as a follower robot, where the DIGIT sensor is attached to the tip of the robot gripper to estimate the contact force. The preliminary results show the capability of using the low-cost vision-based sensor for force feedback applications.

Information-Theoretic Safe Exploration with Gaussian Processes

We consider a sequential decision making task where we are not allowed to evaluate parameters that violate an a priori unknown (safety) constraint. A common approach is to place a Gaussian process prior on the unknown constraint and allow evaluations only in regions that are safe with high probability. Most current methods rely on a discretization of the domain and cannot be directly extended to the continuous case. Moreover, the way in which they exploit regularity assumptions about the constraint introduces an additional critical hyperparameter. In this paper, we propose an information-theoretic safe exploration criterion that directly exploits the GP posterior to identify the most informative safe parameters to evaluate. Our approach is naturally applicable to continuous domains and does not require additional hyperparameters. We theoretically analyze the method and show that we do not violate the safety constraint with high probability and that we explore by learning about the constraint up to arbitrary precision. Empirical evaluations demonstrate improved data-efficiency and scalability.

Hierarchical Policy Blending As Optimal Transport

We present hierarchical policy blending as optimal transport (HiPBOT). This hierarchical framework adapts the weights of low-level reactive expert policies, adding a look-ahead planning layer on the parameter space of a product of expert policies and agents. Our high-level planner realizes a policy blending via unbalanced optimal transport, consolidating the scaling of underlying Riemannian motion policies, effectively adjusting their Riemannian matrix, and deciding over the priorities between experts and agents, guaranteeing safety and task success. Our experimental results in a range of application scenarios from low-dimensional navigation to high-dimensional whole-body control showcase the efficacy and efficiency of HiPBOT, which outperforms state-of-the-art baselines that either perform probabilistic inference or define a tree structure of experts, paving the way for new applications of optimal transport to robot control. More material at https://sites.google.com/view/hipobot

PAC-Bayes Bounds for Bandit Problems: A Survey and Experimental Comparison

PAC-Bayes has recently re-emerged as an effective theory with which one can derive principled learning algorithms with tight performance guarantees. However, applications of PAC-Bayes to bandit problems are relatively rare, which is a great misfortune. Many decision-making problems in healthcare, finance and natural sciences can be modelled as bandit problems. In many of these applications, principled algorithms with strong performance guarantees would be very much appreciated. This survey provides an overview of PAC-Bayes performance bounds for bandit problems and an experimental comparison of these bounds. Our experimental comparison has revealed that available PAC-Bayes upper bounds on the cumulative regret are loose, whereas available PAC-Bayes lower bounds on the expected reward can be surprisingly tight. We found that an offline contextual bandit algorithm that learns a policy by optimising a PAC-Bayes bound was able to learn randomised neural network polices with competitive expected reward and non-vacuous performance guarantees.