We propose Adversarially Trained Actor Critic (ATAC), a new model-free algorithm for offline reinforcement learning (RL) under insufficient data coverage, based on the concept of relative pessimism. ATAC is designed as a two-player Stackelberg game: A policy actor competes against an adversarially trained value critic, who finds data-consistent scenarios where the actor is inferior to the data-collection behavior policy. We prove that, when the actor attains no regret in the two-player game, running ATAC produces a policy that provably 1) outperforms the behavior policy over a wide range of hyperparameters that control the degree of pessimism, and 2) competes with the best policy covered by data with appropriately chosen hyperparameters. Compared with existing works, notably our framework offers both theoretical guarantees for general function approximation and a deep RL implementation scalable to complex environments and large datasets. In the D4RL benchmark, ATAC consistently outperforms state-of-the-art offline RL algorithms on a range of continuous control tasks.
Gradient-based approaches in reinforcement learning (RL) have achieved tremendous success in learning policies for autonomous vehicles. While the performance of these approaches warrants real-world adoption, these policies lack interpretability, limiting deployability in the safety-critical and legally-regulated domain of autonomous driving (AD). AD requires interpretable and verifiable control policies that maintain high performance. We propose Interpretable Continuous Control Trees (ICCTs), a tree-based model that can be optimized via modern, gradient-based, RL approaches to produce high-performing, interpretable policies. The key to our approach is a procedure for allowing direct optimization in a sparse decision-tree-like representation. We validate ICCTs against baselines across six domains, showing that ICCTs are capable of learning interpretable policy representations that parity or outperform baselines by up to 33% in AD scenarios while achieving a 300x-600x reduction in the number of policy parameters against deep learning baselines. Furthermore, we demonstrate the interpretability and utility of our ICCTs through a 14-car physical robot demonstration.
Bisimulation metrics define a distance measure between states of a Markov decision process (MDP) based on a comparison of reward sequences. Due to this property they provide theoretical guarantees in value function approximation (VFA). In this work we first prove that bisimulation and $\pi$-bisimulation metrics can be defined via a more general class of Sinkhorn distances, which unifies various state similarity metrics used in recent work. Then we describe an approximate policy iteration (API) procedure that uses a bisimulation-based discretization of the state space for VFA and prove asymptotic performance bounds. Next, we bound the difference between $\pi$-bisimulation metrics in terms of the change in the policies themselves. Based on these results, we design an API($\alpha$) procedure that employs conservative policy updates and enjoys better performance bounds than the naive API approach. We discuss how such API procedures map onto practical actor-critic methods that use bisimulation metrics for state representation learning. Lastly, we validate our theoretical results and investigate their practical implications via a controlled empirical analysis based on an implementation of bisimulation-based API for finite MDPs.
Distributional reinforcement learning (RL) aims to learn a value-network that predicts the full distribution of the returns for a given state, often modeled via a quantile-based critic. This approach has been successfully integrated into common RL methods for continuous control, giving rise to algorithms such as Distributional Soft Actor-Critic (DSAC). In this paper, we introduce multi-sample target values (MTV) for distributional RL, as a principled replacement for single-sample target value estimation, as commonly employed in current practice. The improved distributional estimates further lend themselves to UCB-based exploration. These two ideas are combined to yield our distributional RL algorithm, E2DC (Extra Exploration with Distributional Critics). We evaluate our approach on a range of continuous control tasks and demonstrate state-of-the-art model-free performance on difficult tasks such as Humanoid control. We provide further insight into the method via visualization and analysis of the learned distributions and their evolution during training.
Many existing imitation learning datasets are collected from multiple demonstrators, each with different expertise at different parts of the environment. Yet, standard imitation learning algorithms typically treat all demonstrators as homogeneous, regardless of their expertise, absorbing the weaknesses of any suboptimal demonstrators. In this work, we show that unsupervised learning over demonstrator expertise can lead to a consistent boost in the performance of imitation learning algorithms. We develop and optimize a joint model over a learned policy and expertise levels of the demonstrators. This enables our model to learn from the optimal behavior and filter out the suboptimal behavior of each demonstrator. Our model learns a single policy that can outperform even the best demonstrator, and can be used to estimate the expertise of any demonstrator at any state. We illustrate our findings on real-robotic continuous control tasks from Robomimic and discrete environments such as MiniGrid and chess, out-performing competing methods in $21$ out of $23$ settings, with an average of $7\%$ and up to $60\%$ improvement in terms of the final reward.
The imitation learning research community has recently made significant progress towards the goal of enabling artificial agents to imitate behaviors from video demonstrations alone. However, current state-of-the-art approaches developed for this problem exhibit high sample complexity due, in part, to the high-dimensional nature of video observations. Towards addressing this issue, we introduce here a new algorithm called Visual Generative Adversarial Imitation from Observation using a State Observer VGAIfO-SO. At its core, VGAIfO-SO seeks to address sample inefficiency using a novel, self-supervised state observer, which provides estimates of lower-dimensional proprioceptive state representations from high-dimensional images. We show experimentally in several continuous control environments that VGAIfO-SO is more sample efficient than other IfO algorithms at learning from video-only demonstrations and can sometimes even achieve performance close to the Generative Adversarial Imitation from Observation (GAIfO) algorithm that has privileged access to the demonstrator's proprioceptive state information.
Proximal Policy Optimization (PPO) methods learn a policy by iteratively performing multiple mini-batch optimization epochs of a surrogate objective with one set of sampled data. Ratio clipping PPO is a popular variant that clips the probability ratios between the target policy and the policy used to collect samples. Ratio clipping yields a pessimistic estimate of the original surrogate objective, and has been shown to be crucial for strong performance. We show in this paper that such ratio clipping may not be a good option as it can fail to effectively bound the ratios. Instead, one can directly optimize the original surrogate objective for multiple epochs; the key is to find a proper condition to early stop the optimization epoch in each iteration. Our theoretical analysis sheds light on how to determine when to stop the optimization epoch, and call the resulting algorithm Early Stopping Policy Optimization (ESPO). We compare ESPO with PPO across many continuous control tasks and show that ESPO significantly outperforms PPO. Furthermore, we show that ESPO can be easily scaled up to distributed training with many workers, delivering strong performance as well.
Application of ensemble of neural networks is becoming an imminent tool for advancing the state-of-the-art in deep reinforcement learning algorithms. However, training these large numbers of neural networks in the ensemble has an exceedingly high computation cost which may become a hindrance in training large-scale systems. In this paper, we propose DNS: a Determinantal Point Process based Neural Network Sampler that specifically uses k-dpp to sample a subset of neural networks for backpropagation at every training step thus significantly reducing the training time and computation cost. We integrated DNS in REDQ for continuous control tasks and evaluated on MuJoCo environments. Our experiments show that DNS augmented REDQ outperforms baseline REDQ in terms of average cumulative reward and achieves this using less than 50% computation when measured in FLOPS.
We consider offline Imitation Learning from corrupted demonstrations where a constant fraction of data can be noise or even arbitrary outliers. Classical approaches such as Behavior Cloning assumes that demonstrations are collected by an presumably optimal expert, hence may fail drastically when learning from corrupted demonstrations. We propose a novel robust algorithm by minimizing a Median-of-Means (MOM) objective which guarantees the accurate estimation of policy, even in the presence of constant fraction of outliers. Our theoretical analysis shows that our robust method in the corrupted setting enjoys nearly the same error scaling and sample complexity guarantees as the classical Behavior Cloning in the expert demonstration setting. Our experiments on continuous-control benchmarks validate that our method exhibits the predicted robustness and effectiveness, and achieves competitive results compared to existing imitation learning methods.
Evolutionary algorithms (EAs) have shown promise in solving sequential decision problems (SDPs) by simplifying them to static optimization problems and searching for the optimal policy parameters in a zeroth-order way. While these methods are highly versatile, they often suffer from high sample complexity due to their ignorance of the underlying temporal structures. In contrast, reinforcement learning (RL) methods typically formulate SDPs as Markov Decision Process (MDP). Although more sample efficient than EAs, RL methods are restricted to differentiable policies and prone to getting stuck in local optima. To address these issues, we propose a novel evolutionary framework Zeroth-Order Actor-Critic (ZOAC). We propose to use step-wise exploration in parameter space and theoretically derive the zeroth-order policy gradient. We further utilize the actor-critic architecture to effectively leverage the Markov property of SDPs and reduce the variance of gradient estimators. In each iteration, ZOAC employs samplers to collect trajectories with parameter space exploration, and alternates between first-order policy evaluation (PEV) and zeroth-order policy improvement (PIM). To evaluate the effectiveness of ZOAC, we apply it to a challenging multi-lane driving task, optimizing the parameters in a rule-based, non-differentiable driving policy that consists of three sub-modules: behavior selection, path planning, and trajectory tracking. We also compare it with gradient-based RL methods on three Gymnasium tasks, optimizing neural network policies with thousands of parameters. Experimental results demonstrate the strong capability of ZOAC in solving SDPs. ZOAC significantly outperforms EAs that treat the problem as static optimization and matches the performance of gradient-based RL methods even without first-order information, in terms of total average return across all tasks.
Model-free continuous control for robot navigation tasks using Deep Reinforcement Learning (DRL) that relies on noisy policies for exploration is sensitive to the density of rewards. In practice, robots are usually deployed in cluttered environments, containing many obstacles and narrow passageways. Designing dense effective rewards is challenging, resulting in exploration issues during training. Such a problem becomes even more serious when tasks are described using temporal logic specifications. This work presents a deep policy gradient algorithm for controlling a robot with unknown dynamics operating in a cluttered environment when the task is specified as a Linear Temporal Logic (LTL) formula. To overcome the environmental challenge of exploration during training, we propose a novel path planning-guided reward scheme by integrating sampling-based methods to effectively complete goal-reaching missions. To facilitate LTL satisfaction, our approach decomposes the LTL mission into sub-goal-reaching tasks that are solved in a distributed manner. Our framework is shown to significantly improve performance (effectiveness, efficiency) and exploration of robots tasked with complex missions in large-scale cluttered environments. A video demonstration can be found on YouTube Channel: https://youtu.be/yMh_NUNWxho.
Hierarchical reinforcement learning (HRL) proposes to solve difficult tasks by performing decision-making and control at successively higher levels of temporal abstraction. However, off-policy HRL often suffers from the problem of a non-stationary high-level policy since the low-level policy is constantly changing. In this paper, we propose a novel HRL approach for mitigating the non-stationarity by adversarially enforcing the high-level policy to generate subgoals compatible with the current instantiation of the low-level policy. In practice, the adversarial learning is implemented by training a simple state-conditioned discriminator network concurrently with the high-level policy which determines the compatibility level of subgoals. Comparison to state-of-the-art algorithms shows that our approach improves both learning efficiency and performance in challenging continuous control tasks.
As an important algorithm in deep reinforcement learning, advantage actor critic (A2C) has been widely succeeded in both discrete and continuous control tasks with raw pixel inputs, but its sample efficiency still needs to improve more. In traditional reinforcement learning, actor-critic algorithms generally use the recursive least squares (RLS) technology to update the parameter of linear function approximators for accelerating their convergence speed. However, A2C algorithms seldom use this technology to train deep neural networks (DNNs) for improving their sample efficiency. In this paper, we propose two novel RLS-based A2C algorithms and investigate their performance. Both proposed algorithms, called RLSSA2C and RLSNA2C, use the RLS method to train the critic network and the hidden layers of the actor network. The main difference between them is at the policy learning step. RLSSA2C uses an ordinary first-order gradient descent algorithm and the standard policy gradient to learn the policy parameter. RLSNA2C uses the Kronecker-factored approximation, the RLS method and the natural policy gradient to learn the compatible parameter and the policy parameter. In addition, we analyze the complexity and convergence of both algorithms, and present three tricks for further improving their convergence speed. Finally, we demonstrate the effectiveness of both algorithms on 40 games in the Atari 2600 environment and 11 tasks in the MuJoCo environment. From the experimental results, it is shown that our both algorithms have better sample efficiency than the vanilla A2C on most games or tasks, and have higher computational efficiency than other two state-of-the-art algorithms.
Evolutionary Algorithms (EAs) and Deep Reinforcement Learning (DRL) have recently been integrated to take the advantage of the both methods for better exploration and exploitation.The evolutionary part in these hybrid methods maintains a population of policy networks.However, existing methods focus on optimizing the parameters of policy network, which is usually high-dimensional and tricky for EA.In this paper, we shift the target of evolution from high-dimensional parameter space to low-dimensional action space.We propose Evolutionary Action Selection-Twin Delayed Deep Deterministic Policy Gradient (EAS-TD3), a novel hybrid method of EA and DRL.In EAS, we focus on optimizing the action chosen by the policy network and attempt to obtain high-quality actions to promote policy learning through an evolutionary algorithm. We conduct several experiments on challenging continuous control tasks.The result shows that EAS-TD3 shows superior performance over other state-of-art methods.
In model-free deep reinforcement learning (RL) algorithms, using noisy value estimates to supervise policy evaluation and optimization is detrimental to the sample efficiency. As this noise is heteroscedastic, its effects can be mitigated using uncertainty-based weights in the optimization process. Previous methods rely on sampled ensembles, which do not capture all aspects of uncertainty. We provide a systematic analysis of the sources of uncertainty in the noisy supervision that occurs in RL, and introduce inverse-variance RL, a Bayesian framework which combines probabilistic ensembles and Batch Inverse Variance weighting. We propose a method whereby two complementary uncertainty estimation methods account for both the Q-value and the environment stochasticity to better mitigate the negative impacts of noisy supervision. Our results show significant improvement in terms of sample efficiency on discrete and continuous control tasks.
The integration of Reinforcement Learning (RL) and Evolutionary Algorithms (EAs) aims at simultaneously exploiting the sample efficiency as well as the diversity and robustness of the two paradigms. Recently, hybrid learning frameworks based on this principle have achieved great success in various challenging robot control tasks. However, in these methods, policies from the genetic population are evaluated via interactions with the real environments, limiting their applicability in computationally expensive problems. In this work, we propose Surrogate-assisted Controller (SC), a novel and efficient module that can be integrated into existing frameworks to alleviate the computational burden of EAs by partially replacing the expensive policy evaluation. The key challenge in applying this module is to prevent the optimization process from being misled by the possible false minima introduced by the surrogate. To address this issue, we present two strategies for SC to control the workflow of hybrid frameworks. Experiments on six continuous control tasks from the OpenAI Gym platform show that SC can not only significantly reduce the cost of fitness evaluations, but also boost the performance of the original hybrid frameworks with collaborative learning and evolutionary processes.
Deep Reinforcement Learning (RL) is a powerful framework for solving complex real-world problems. Large neural networks employed in the framework are traditionally associated with better generalization capabilities, but their increased size entails the drawbacks of extensive training duration, substantial hardware resources, and longer inference times. One way to tackle this problem is to prune neural networks leaving only the necessary parameters. State-of-the-art concurrent pruning techniques for imposing sparsity perform demonstrably well in applications where data distributions are fixed. However, they have not yet been substantially explored in the context of RL. We close the gap between RL and single-shot pruning techniques and present a general pruning approach to the Offline RL. We leverage a fixed dataset to prune neural networks before the start of RL training. We then run experiments varying the network sparsity level and evaluating the validity of pruning at initialization techniques in continuous control tasks. Our results show that with 95% of the network weights pruned, Offline-RL algorithms can still retain performance in the majority of our experiments. To the best of our knowledge, no prior work utilizing pruning in RL retained performance at such high levels of sparsity. Moreover, pruning at initialization techniques can be easily integrated into any existing Offline-RL algorithms without changing the learning objective.
Although it is well known that exploration plays a key role in Reinforcement Learning (RL), prevailing exploration strategies for continuous control tasks in RL are mainly based on naive isotropic Gaussian noise regardless of the causality relationship between action space and the task and consider all dimensions of actions equally important. In this work, we propose to conduct interventions on the primal action space to discover the causal relationship between the action space and the task reward. We propose the method of State-Wise Action Refined (SWAR), which addresses the issue of action space redundancy and promote causality discovery in RL. We formulate causality discovery in RL tasks as a state-dependent action space selection problem and propose two practical algorithms as solutions. The first approach, TD-SWAR, detects task-related actions during temporal difference learning, while the second approach, Dyn-SWAR, reveals important actions through dynamic model prediction. Empirically, both methods provide approaches to understand the decisions made by RL agents and improve learning efficiency in action-redundant tasks.
Quantum computing has a superior advantage in tackling specific problems, such as integer factorization and Simon's problem. For more general tasks in machine learning, by applying variational quantum circuits, more and more quantum algorithms have been proposed recently, especially in supervised learning and unsupervised learning. However, little work has been done in reinforcement learning, arguably more important and challenging. Previous work in quantum reinforcement learning mainly focuses on discrete control tasks where the action space is discrete. In this work, we develop a quantum reinforcement learning algorithm based on soft actor-critic -- one of the state-of-the-art methods for continuous control. Specifically, we use a hybrid quantum-classical policy network consisting of a variational quantum circuit and a classical artificial neural network. Tested in a standard reinforcement learning benchmark, we show that this quantum version of soft actor-critic is comparable with the original soft actor-critic, using much less adjustable parameters. Furthermore, we analyze the effect of different hyper-parameters and policy network architectures, pointing out the importance of architecture design for quantum reinforcement learning.
Deep reinforcement learning (RL) has recently shown great promise in robotic continuous control tasks. Nevertheless, prior research in this vein center around the centralized learning setting that largely relies on the communication availability among all the components of a robot. However, agents in the real world often operate in a decentralised fashion without communication due to latency requirements, limited power budgets and safety concerns. By formulating robotic components as a system of decentralised agents, this work presents a decentralised multiagent reinforcement learning framework for continuous control. To this end, we first develop a cooperative multiagent PPO framework that allows for centralized optimisation during training and decentralised operation during execution. However, the system only receives a global reward signal which is not attributed towards each agent. To address this challenge, we further propose a generic game-theoretic credit assignment framework which computes agent-specific reward signals. Last but not least, we also incorporate a model-based RL module into our credit assignment framework, which leads to significant improvement in sample efficiency. We demonstrate the effectiveness of our framework on experimental results on Mujoco locomotion control tasks. For a demo video please visit: https://youtu.be/gFyVPm4svEY.