Offline reinforcement learning has shown great promise in leveraging large pre-collected datasets for policy learning, allowing agents to forgo often-expensive online data collection. However, offline reinforcement learning from visual observations with continuous action spaces remains under-explored, with a limited understanding of the key challenges in this complex domain. In this paper, we establish simple baselines for continuous control in the visual domain and introduce a suite of benchmarking tasks for offline reinforcement learning from visual observations designed to better represent the data distributions present in real-world offline RL problems and guided by a set of desiderata for offline RL from visual observations, including robustness to visual distractions and visually identifiable changes in dynamics. Using this suite of benchmarking tasks, we show that simple modifications to two popular vision-based online reinforcement learning algorithms, DreamerV2 and DrQ-v2, suffice to outperform existing offline RL methods and establish competitive baselines for continuous control in the visual domain. We rigorously evaluate these algorithms and perform an empirical evaluation of the differences between state-of-the-art model-based and model-free offline RL methods for continuous control from visual observations. All code and data used in this evaluation are open-sourced to facilitate progress in this domain.

Standard reinforcement learning (RL) agents never intelligently explore like a human (i.e. by taking into account complex domain priors and previous explorations). Even the most basic intelligent exploration strategies such as exhaustive search are only inefficiently or poorly approximated by approaches such as novelty search or intrinsic motivation, let alone more complicated strategies like learning new skills, climbing stairs, opening doors, or conducting experiments. This lack of intelligent exploration limits sample efficiency and prevents solving hard exploration domains. We argue a core barrier prohibiting many RL approaches from learning intelligent exploration is that the methods attempt to explore and exploit simultaneously, which harms both exploration and exploitation as the goals often conflict. We propose a novel meta-RL framework (First-Explore) with two policies: one policy learns to only explore and one policy learns to only exploit. Once trained, we can then explore with the explore policy, for as long as desired, and then exploit based on all the information gained during exploration. This approach avoids the conflict of trying to do both exploration and exploitation at once. We demonstrate that First-Explore can learn intelligent exploration strategies such as exhaustive search and more, and that it outperforms dominant standard RL and meta-RL approaches on domains where exploration requires sacrificing reward. First-Explore is a significant step towards creating meta-RL algorithms capable of learning human-level exploration which is essential to solve challenging unseen hard-exploration domains.

Mobile edge computing (MEC) is essential for next-generation mobile network applications that prioritize various performance metrics, including delays and energy consumption. However, conventional single-objective scheduling solutions cannot be directly applied to practical systems in which the preferences of these applications (i.e., the weights of different objectives) are often unknown or challenging to specify in advance. In this study, we address this issue by formulating a multi-objective offloading problem for MEC with multiple edges to minimize expected long-term energy consumption and transmission delay while considering unknown preferences as parameters. To address the challenge of unknown preferences, we design a multi-objective (deep) reinforcement learning (MORL)-based resource scheduling scheme with proximal policy optimization (PPO). In addition, we introduce a well-designed state encoding method for constructing features for multiple edges in MEC systems, a sophisticated reward function for accurately computing the utilities of delay and energy consumption. Simulation results demonstrate that our proposed MORL scheme enhances the hypervolume of the Pareto front by up to 233.1% compared to benchmarks. Our full framework is available at https://github.com/gracefulning/mec_morl_multipolicy.

Text-based reinforcement learning agents have predominantly been neural network-based models with embeddings-based representation, learning uninterpretable policies that often do not generalize well to unseen games. On the other hand, neuro-symbolic methods, specifically those that leverage an intermediate formal representation, are gaining significant attention in language understanding tasks. This is because of their advantages ranging from inherent interpretability, the lesser requirement of training data, and being generalizable in scenarios with unseen data. Therefore, in this paper, we propose a modular, NEuro-Symbolic Textual Agent (NESTA) that combines a generic semantic parser with a rule induction system to learn abstract interpretable rules as policies. Our experiments on established text-based game benchmarks show that the proposed NESTA method outperforms deep reinforcement learning-based techniques by achieving better generalization to unseen test games and learning from fewer training interactions.

Large events such as conferences, concerts and sports games, often cause surges in demand for ride services that are not captured in average demand patterns, posing unique challenges for routing algorithms. We propose a learning framework for an autonomous fleet of taxis that scrapes event data from the internet to predict and adapt to surges in demand and generates cooperative routing and pickup policies that service a higher number of requests than other routing protocols. We achieve this through a combination of (i) an event processing framework that scrapes the internet for event information and generates dense vector representations that can be used as input features for a neural network that predicts demand; (ii) a two neural network system that predicts hourly demand over the entire map, using these dense vector representations; (iii) a probabilistic approach that leverages locale occupancy schedules to map publicly available demand data over sectors to discretized street intersections; and finally, (iv) a scalable model-based reinforcement learning framework that uses the predicted demand over intersections to anticipate surges and route taxis using one-agent-at-a-time rollout with limited sampling certainty equivalence. We learn routing and pickup policies using real NYC ride share data for 2022 and information for more than 2000 events across 300 unique venues in Manhattan. We test our approach with a fleet of 100 taxis on a map with 38 different sectors (2235 street intersections). Our experimental results demonstrate that our method obtains routing policies that service $6$ more requests on average per minute (around $360$ more requests per hour) than other model-based RL frameworks and other classical algorithms in operations research when dealing with surge demand conditions.

The field of cooperative multi-agent reinforcement learning (MARL) has seen widespread use in addressing complex coordination tasks. While value decomposition methods in MARL have been popular, they have limitations in solving tasks with non-monotonic returns, restricting their general application. Our work highlights the significance of joint intentions in cooperation, which can overcome non-monotonic problems and increase the interpretability of the learning process. To this end, we present a novel MARL method that leverages learnable joint intentions. Our method employs a hierarchical framework consisting of a joint intention policy and a behavior policy to formulate the optimal cooperative policy. The joint intentions are autonomously learned in a latent space through unsupervised learning and enable the method adaptable to different agent configurations. Our results demonstrate significant performance improvements in both the StarCraft micromanagement benchmark and challenging MAgent domains, showcasing the effectiveness of our method in learning meaningful joint intentions.

Given a limited labeling budget, active learning (al) aims to sample the most informative instances from an unlabeled pool to acquire labels for subsequent model training. To achieve this, al typically measures the informativeness of unlabeled instances based on uncertainty and diversity. However, it does not consider erroneous instances with their neighborhood error density, which have great potential to improve the model performance. To address this limitation, we propose Real, a novel approach to select data instances with Representative Errors for Active Learning. It identifies minority predictions as pseudo errors within a cluster and allocates an adaptive sampling budget for the cluster based on estimated error density. Extensive experiments on five text classification datasets demonstrate that Real consistently outperforms all best-performing baselines regarding accuracy and F1-macro scores across a wide range of hyperparameter settings. Our analysis also shows that Real selects the most representative pseudo errors that match the distribution of ground-truth errors along the decision boundary.

The ever-growing complexity of reinforcement learning (RL) tasks demands a distributed RL system to efficiently generate and process a massive amount of data to train intelligent agents. However, existing open-source libraries suffer from various limitations, which impede their practical use in challenging scenarios where large-scale training is necessary. While industrial systems from OpenAI and DeepMind have achieved successful large-scale RL training, their system architecture and implementation details remain undisclosed to the community. In this paper, we present a novel abstraction on the dataflows of RL training, which unifies practical RL training across diverse applications into a general framework and enables fine-grained optimizations. Following this abstraction, we develop a scalable, efficient, and extensible distributed RL system called ReaLly Scalable RL (SRL). The system architecture of SRL separates major RL computation components and allows massively parallelized training. Moreover, SRL offers user-friendly and extensible interfaces for customized algorithms. Our evaluation shows that SRL outperforms existing academic libraries in both a single machine and a medium-sized cluster. In a large-scale cluster, the novel architecture of SRL leads to up to 3.7x speedup compared to the design choices adopted by the existing libraries. We also conduct a direct benchmark comparison to OpenAI's industrial system, Rapid, in the challenging hide-and-seek environment. SRL reproduces the same solution as reported by OpenAI with up to 5x speedup in wall-clock time. Furthermore, we also examine the performance of SRL in a much harder variant of the hide-and-seek environment and achieve substantial learning speedup by scaling SRL to over 15k CPU cores and 32 A100 GPUs. Notably, SRL is the first in the academic community to perform RL experiments at such a large scale.

Reinforcement learning (RL) has made significant strides in various complex domains. However, identifying an effective policy via RL often necessitates extensive exploration. Imitation learning aims to mitigate this issue by using expert demonstrations to guide exploration. In real-world scenarios, one often has access to multiple suboptimal black-box experts, rather than a single optimal oracle. These experts do not universally outperform each other across all states, presenting a challenge in actively deciding which oracle to use and in which state. We introduce MAPS and MAPS-SE, a class of policy improvement algorithms that perform imitation learning from multiple suboptimal oracles. In particular, MAPS actively selects which of the oracles to imitate and improve their value function estimates, and MAPS-SE additionally leverages an active state exploration criterion to determine which states one should explore. We provide a comprehensive theoretical analysis and demonstrate that MAPS and MAPS-SE enjoy sample efficiency advantage over the state-of-the-art policy improvement algorithms. Empirical results show that MAPS-SE significantly accelerates policy optimization via state-wise imitation learning from multiple oracles across a broad spectrum of control tasks in the DeepMind Control Suite. Our code is publicly available at: https://github.com/ripl/maps.

Originally developed for natural language problems, transformer models have recently been widely used in offline reinforcement learning tasks. This is because the agent's history can be represented as a sequence, and the whole task can be reduced to the sequence modeling task. However, the quadratic complexity of the transformer operation limits the potential increase in context. Therefore, different versions of the memory mechanism are used to work with long sequences in a natural language. This paper proposes the Recurrent Memory Decision Transformer (RMDT), a model that uses a recurrent memory mechanism for reinforcement learning problems. We conduct thorough experiments on Atari games and MuJoCo control problems and show that our proposed model is significantly superior to its counterparts without the recurrent memory mechanism on Atari games. We also carefully study the effect of memory on the performance of the proposed model. These findings shed light on the potential of incorporating recurrent memory mechanisms to improve the performance of large-scale transformer models in offline reinforcement learning tasks. The Recurrent Memory Decision Transformer code is publicly available in the repository \url{https://anonymous.4open.science/r/RMDT-4FE4}.