Offline Goal-Conditioned Reinforcement Learning (GCRL) is tasked with learning to achieve multiple goals in an environment purely from offline datasets using sparse reward functions. Offline GCRL is pivotal for developing generalist agents capable of leveraging pre-existing datasets to learn diverse and reusable skills without hand-engineering reward functions. However, contemporary approaches to GCRL based on supervised learning and contrastive learning are often suboptimal in the offline setting. An alternative perspective on GCRL optimizes for occupancy matching, but necessitates learning a discriminator, which subsequently serves as a pseudo-reward for downstream RL. Inaccuracies in the learned discriminator can cascade, negatively influencing the resulting policy. We present a novel approach to GCRL under a new lens of mixture-distribution matching, leading to our discriminator-free method: SMORe. The key insight is combining the occupancy matching perspective of GCRL with a convex dual formulation to derive a learning objective that can better leverage suboptimal offline data. SMORe learns scores or unnormalized densities representing the importance of taking an action at a state for reaching a particular goal. SMORe is principled and our extensive experiments on the fully offline GCRL benchmark composed of robot manipulation and locomotion tasks, including high-dimensional observations, show that SMORe can outperform state-of-the-art baselines by a significant margin.
Reinforcement Learning from Human Feedback (RLHF) has emerged as a popular paradigm for aligning models with human intent. Typically RLHF algorithms operate in two phases: first, use human preferences to learn a reward function and second, align the model by optimizing the learned reward via reinforcement learning (RL). This paradigm assumes that human preferences are distributed according to reward, but recent work suggests that they instead follow the regret under the user's optimal policy. Thus, learning a reward function from feedback is not only based on a flawed assumption of human preference, but also leads to unwieldy optimization challenges that stem from policy gradients or bootstrapping in the RL phase. Because of these optimization challenges, contemporary RLHF methods restrict themselves to contextual bandit settings (e.g., as in large language models) or limit observation dimensionality (e.g., state-based robotics). We overcome these limitations by introducing a new family of algorithms for optimizing behavior from human feedback using the regret-based model of human preferences. Using the principle of maximum entropy, we derive Contrastive Preference Learning (CPL), an algorithm for learning optimal policies from preferences without learning reward functions, circumventing the need for RL. CPL is fully off-policy, uses only a simple contrastive objective, and can be applied to arbitrary MDPs. This enables CPL to elegantly scale to high-dimensional and sequential RLHF problems while being simpler than prior methods.
Reinforcement Learning from Human Feedback (RLHF) has emerged as a popular paradigm for aligning models with human intent. Typically RLHF algorithms operate in two phases: first, use human preferences to learn a reward function and second, align the model by optimizing the learned reward via reinforcement learning (RL). This paradigm assumes that human preferences are distributed according to reward, but recent work suggests that they instead follow the regret under the user's optimal policy. Thus, learning a reward function from feedback is not only based on a flawed assumption of human preference, but also leads to unwieldy optimization challenges that stem from policy gradients or bootstrapping in the RL phase. Because of these optimization challenges, contemporary RLHF methods restrict themselves to contextual bandit settings (e.g., as in large language models) or limit observation dimensionality (e.g., state-based robotics). We overcome these limitations by introducing a new family of algorithms for optimizing behavior from human feedback using the regret-based model of human preferences. Using the principle of maximum entropy, we derive Contrastive Preference Learning (CPL), an algorithm for learning optimal policies from preferences without learning reward functions, circumventing the need for RL. CPL is fully off-policy, uses only a simple contrastive objective, and can be applied to arbitrary MDPs. This enables CPL to elegantly scale to high-dimensional and sequential RLHF problems while being simpler than prior methods.
It is well known that Reinforcement Learning (RL) can be formulated as a convex program with linear constraints. The dual form of this formulation is unconstrained, which we refer to as dual RL, and can leverage preexisting tools from convex optimization to improve the learning performance of RL agents. We show that several state-of-the-art deep RL algorithms (in online, offline, and imitation settings) can be viewed as dual RL approaches in a unified framework. This unification calls for the methods to be studied on common ground, so as to identify the components that actually contribute to the success of these methods. Our unification also reveals that prior off-policy imitation learning methods in the dual space are based on an unrealistic coverage assumption and are restricted to matching a particular f-divergence. We propose a new method using a simple modification to the dual framework that allows for imitation learning with arbitrary off-policy data to obtain near-expert performance.
We present our runner-up approach for the Real Robot Challenge 2021. We build upon our previous approach used in Real Robot Challenge 2020. To solve the task of sequential goal-reaching we focus on two aspects to achieving near-optimal trajectory: Grasp stability and Controller performance. In the RRC 2021 simulated challenge, our method relied on a hand-designed Pinch grasp combined with Trajectory Interpolation for better stability during the motion for fast goal-reaching. In Stage 1, we observe reverting to a Triangular grasp to provide a more stable grasp when combined with Trajectory Interpolation, possibly due to the sim2real gap. The video demonstration for our approach is available at https://youtu.be/dlOueoaRWrM. The code is publicly available at https://github.com/madan96/benchmark-rrc.
We propose a new framework for imitation learning - treating imitation as a two-player ranking-based Stackelberg game between a $\textit{policy}$ and a $\textit{reward}$ function. In this game, the reward agent learns to satisfy pairwise performance rankings within a set of policies, while the policy agent learns to maximize this reward. This game encompasses a large subset of both inverse reinforcement learning (IRL) methods and methods which learn from offline preferences. The Stackelberg game formulation allows us to use optimization methods that take the game structure into account, leading to more sample efficient and stable learning dynamics compared to existing IRL methods. We theoretically analyze the requirements of the loss function used for ranking policy performances to facilitate near-optimal imitation learning at equilibrium. We use insights from this analysis to further increase sample efficiency of the ranking game by using automatically generated rankings or with offline annotated rankings. Our experiments show that the proposed method achieves state-of-the-art sample efficiency and is able to solve previously unsolvable tasks in the Learning from Observation (LfO) setting.
Deploying Reinforcement Learning (RL) agents in the real-world require that the agents satisfy safety constraints. Current RL agents explore the environment without considering these constraints, which can lead to damage to the hardware or even other agents in the environment. We propose a new method, LBPO, that uses a Lyapunov-based barrier function to restrict the policy update to a safe set for each training iteration. Our method also allows the user to control the conservativeness of the agent with respect to the constraints in the environment. LBPO significantly outperforms state-of-the-art baselines in terms of the number of constraint violations during training while being competitive in terms of performance. Further, our analysis reveals that baselines like CPO and SDDPG rely mostly on backtracking to ensure safety rather than safe projection, which provides insight into why previous methods might not have effectively limit the number of constraint violations.
Imitation learning is well-suited for robotic tasks where it is difficult to directly program the behavior or specify a cost for optimal control. In this work, we propose a method for learning the reward function (and the corresponding policy) to match the expert state density. Our main result is the analytic gradient of any f-divergence between the agent and expert state distribution w.r.t. reward parameters. Based on the derived gradient, we present an algorithm, f-IRL, that recovers a stationary reward function from the expert density by gradient descent. We show that f-IRL can learn behaviors from a hand-designed target state density or implicitly through expert observations. Our method outperforms adversarial imitation learning methods in terms of sample efficiency and the required number of expert trajectories on IRL benchmarks. Moreover, we show that the recovered reward function can be used to quickly solve downstream tasks, and empirically demonstrate its utility on hard-to-explore tasks and for behavior transfer across changes in dynamics.
We explore the application of normalizing flows for improving the performance of trajectory planning for autonomous vehicles (AVs). Normalizing flows provide an invertible mapping from a known prior distribution to a potentially complex, multi-modal target distribution and allow for fast sampling with exact PDF inference. By modeling a trajectory planner's cost manifold as an energy function we learn a scene conditioned mapping from the prior to a Boltzmann distribution over the AV control space. This mapping allows for control samples and their associated energy to be generated jointly and in parallel. We propose using neural autoregressive flow (NAF) as part of an end-to-end deep learned system that allows for utilizing sensors, map, and route information to condition the flow mapping. Finally, we demonstrate the effectiveness of our approach on real world datasets over IL and hand constructed trajectory sampling techniques.
We propose Learning Off-Policy with Online Planning (LOOP), combining the techniques from model-based and model-free reinforcement learning algorithms. The agent learns a model of the environment, and then uses trajectory optimization with the learned model to select actions. To sidestep the myopic effect of fixed horizon trajectory optimization, a value function is attached to the end of the planning horizon. This value function is learned through off-policy reinforcement learning, using trajectory optimization as its behavior policy. Furthermore, we introduce "actor-guided" trajectory optimization to mitigate the actor-divergence issue in the proposed method. We benchmark our methods on continuous control tasks and demonstrate that it offers a significant improvement over the underlying model-based and model-free algorithms.