Offline reinforcement learning (RL) provides a promising direction to exploit the massive amount of offline data for complex decision-making tasks. Due to the distribution shift issue, current offline RL algorithms are generally designed to be conservative for value estimation and action selection. However, such conservatism impairs the robustness of learned policies, leading to a significant change even for a small perturbation on observations. To trade off robustness and conservatism, we propose Robust Offline Reinforcement Learning (RORL) with a novel conservative smoothing technique. In RORL, we explicitly introduce regularization on the policy and the value function for states near the dataset and additional conservative value estimation on these OOD states. Theoretically, we show RORL enjoys a tighter suboptimality bound than recent theoretical results in linear MDPs. We demonstrate that RORL can achieve the state-of-the-art performance on the general offline RL benchmark and is considerably robust to adversarial observation perturbation.
Solving goal-conditioned tasks with sparse rewards using self-supervised learning is promising because of its simplicity and stability over current reinforcement learning (RL) algorithms. A recent work, called Goal-Conditioned Supervised Learning (GCSL), provides a new learning framework by iteratively relabeling and imitating self-generated experiences. In this paper, we revisit the theoretical property of GCSL -- optimizing a lower bound of the goal reaching objective, and extend GCSL as a novel offline goal-conditioned RL algorithm. The proposed method is named Weighted GCSL (WGCSL), in which we introduce an advanced compound weight consisting of three parts (1) discounted weight for goal relabeling, (2) goal-conditioned exponential advantage weight, and (3) best-advantage weight. Theoretically, WGCSL is proved to optimize an equivalent lower bound of the goal-conditioned RL objective and generates monotonically improved policies via an iterated scheme. The monotonic property holds for any behavior policies, and therefore WGCSL can be applied to both online and offline settings. To evaluate algorithms in the offline goal-conditioned RL setting, we provide a benchmark including a range of point and simulated robot domains. Experiments in the introduced benchmark demonstrate that WGCSL can consistently outperform GCSL and existing state-of-the-art offline methods in the fully offline goal-conditioned setting.
Exploration methods based on pseudo-count of transitions or curiosity of dynamics have achieved promising results in solving reinforcement learning with sparse rewards. However, such methods are usually sensitive to environmental dynamics-irrelevant information, e.g., white-noise. To handle such dynamics-irrelevant information, we propose a Dynamic Bottleneck (DB) model, which attains a dynamics-relevant representation based on the information-bottleneck principle. Based on the DB model, we further propose DB-bonus, which encourages the agent to explore state-action pairs with high information gain. We establish theoretical connections between the proposed DB-bonus, the upper confidence bound (UCB) for linear case, and the visiting count for tabular case. We evaluate the proposed method on Atari suits with dynamics-irrelevant noises. Our experiments show that exploration with DB bonus outperforms several state-of-the-art exploration methods in noisy environments.
In multi-agent reinforcement learning, the behaviors that agents learn in a single Markov Game (MG) are typically confined to the given agent number (i.e., population size). Every single MG induced by varying population sizes may possess distinct optimal joint strategies and game-specific knowledge, which are modeled independently in modern multi-agent algorithms. In this work, we focus on creating agents that generalize across population-varying MGs. Instead of learning a unimodal policy, each agent learns a policy set that is formed by effective strategies across a variety of games. We propose Meta Representations for Agents (MRA) that explicitly models the game-common and game-specific strategic knowledge. By representing the policy sets with multi-modal latent policies, the common strategic knowledge and diverse strategic modes are discovered with an iterative optimization procedure. We prove that as an approximation to a constrained mutual information maximization objective, the learned policies can reach Nash Equilibrium in every evaluation MG under the assumption of Lipschitz game on a sufficiently large latent space. When deploying it at practical latent models with limited size, fast adaptation can be achieved by leveraging the first-order gradient information. Extensive experiments show the effectiveness of MRA on both training performance and generalization ability in hard and unseen games.
Solving multi-goal reinforcement learning (RL) problems with sparse rewards is generally challenging. Existing approaches have utilized goal relabeling on collected experiences to alleviate issues raised from sparse rewards. However, these methods are still limited in efficiency and cannot make full use of experiences. In this paper, we propose Model-based Hindsight Experience Replay (MHER), which exploits experiences more efficiently by leveraging environmental dynamics to generate virtual achieved goals. Replacing original goals with virtual goals generated from interaction with a trained dynamics model leads to a novel relabeling method, \emph{model-based relabeling} (MBR). Based on MBR, MHER performs both reinforcement learning and supervised learning for efficient policy improvement. Theoretically, we also prove the supervised part in MHER, i.e., goal-conditioned supervised learning with MBR data, optimizes a lower bound on the multi-goal RL objective. Experimental results in several point-based tasks and simulated robotics environments show that MHER achieves significantly higher sample efficiency than previous state-of-the-art methods.
One principled approach for provably efficient exploration is incorporating the upper confidence bound (UCB) into the value function as a bonus. However, UCB is specified to deal with linear and tabular settings and is incompatible with Deep Reinforcement Learning (DRL). In this paper, we propose a principled exploration method for DRL through Optimistic Bootstrapping and Backward Induction (OB2I). OB2I constructs a general-purpose UCB-bonus through non-parametric bootstrap in DRL. The UCB-bonus estimates the epistemic uncertainty of state-action pairs for optimistic exploration. We build theoretical connections between the proposed UCB-bonus and the LSVI-UCB in a linear setting. We propagate future uncertainty in a time-consistent manner through episodic backward update, which exploits the theoretical advantage and empirically improves the sample-efficiency. Our experiments in the MNIST maze and Atari suite suggest that OB2I outperforms several state-of-the-art exploration approaches.
Competitive Self-Play (CSP) based Multi-Agent Reinforcement Learning (MARL) has shown phenomenal breakthroughs recently. Strong AIs are achieved for several benchmarks, including Dota 2, Glory of Kings, Quake III, StarCraft II, to name a few. Despite the success, the MARL training is extremely data thirsty, requiring typically billions of (if not trillions of) frames be seen from the environment during training in order for learning a high performance agent. This poses non-trivial difficulties for researchers or engineers and prevents the application of MARL to a broader range of real-world problems. To address this issue, in this manuscript we describe a framework, referred to as TLeague, that aims at large-scale training and implements several main-stream CSP-MARL algorithms. The training can be deployed in either a single machine or a cluster of hybrid machines (CPUs and GPUs), where the standard Kubernetes is supported in a cloud native manner. TLeague achieves a high throughput and a reasonable scale-up when performing distributed training. Thanks to the modular design, it is also easy to extend for solving other multi-agent problems or implementing and verifying MARL algorithms. We present experiments over StarCraft II, ViZDoom and Pommerman to show the efficiency and effectiveness of TLeague. The code is open-sourced and available at https://github.com/tencent-ailab/tleague_projpage
StarCraft, one of the most difficult esport games with long-standing history of professional tournaments, has attracted generations of players and fans, and also, intense attentions in artificial intelligence research. Recently, Google's DeepMind announced AlphaStar, a grandmaster level AI in StarCraft II. In this paper, we introduce a new AI agent, named TStarBot-X, that is trained under limited computation resources and can play competitively with expert human players. TStarBot-X takes advantage of important techniques introduced in AlphaStar, and also benefits from substantial innovations including new league training methods, novel multi-agent roles, rule-guided policy search, lightweight neural network architecture, and importance sampling in imitation learning, etc. We show that with limited computation resources, a faithful reimplementation of AlphaStar can not succeed and the proposed techniques are necessary to ensure TStarBot-X's competitive performance. We reveal all technical details that are complementary to those mentioned in AlphaStar, showing the most sensitive parts in league training, reinforcement learning and imitation learning that affect the performance of the agents. Most importantly, this is an open-sourced study that all codes and resources (including the trained model parameters) are publicly accessible via https://github.com/tencent-ailab/tleague_projpage We expect this study could be beneficial for both academic and industrial future research in solving complex problems like StarCraft, and also, might provide a sparring partner for all StarCraft II players and other AI agents.