Learning on One Mode: Addressing Multi-Modality in Offline Reinforcement Learning

Offline reinforcement learning (RL) seeks to learn optimal policies from static datasets without interacting with the environment. A common challenge is handling multi-modal action distributions, where multiple behaviours are represented in the data. Existing methods often assume unimodal behaviour policies, leading to suboptimal performance when this assumption is violated. We propose Weighted Imitation Learning on One Mode (LOM), a novel approach that focuses on learning from a single, promising mode of the behaviour policy. By using a Gaussian mixture model to identify modes and selecting the best mode based on expected returns, LOM avoids the pitfalls of averaging over conflicting actions. Theoretically, we show that LOM improves performance while maintaining simplicity in policy learning. Empirically, LOM outperforms existing methods on standard D4RL benchmarks and demonstrates its effectiveness in complex, multi-modal scenarios.

An Investigation of Offline Reinforcement Learning in Factorisable Action Spaces

Expanding reinforcement learning (RL) to offline domains generates promising prospects, particularly in sectors where data collection poses substantial challenges or risks. Pivotal to the success of transferring RL offline is mitigating overestimation bias in value estimates for state-action pairs absent from data. Whilst numerous approaches have been proposed in recent years, these tend to focus primarily on continuous or small-scale discrete action spaces. Factorised discrete action spaces, on the other hand, have received relatively little attention, despite many real-world problems naturally having factorisable actions. In this work, we undertake a formative investigation into offline reinforcement learning in factorisable action spaces. Using value-decomposition as formulated in DecQN as a foundation, we present the case for a factorised approach and conduct an extensive empirical evaluation of several offline techniques adapted to the factorised setting. In the absence of established benchmarks, we introduce a suite of our own comprising datasets of varying quality and task complexity. Advocating for reproducible research and innovation, we make all datasets available for public use alongside our code base.

Mitigating Relative Over-Generalization in Multi-Agent Reinforcement Learning

In decentralized multi-agent reinforcement learning, agents learning in isolation can lead to relative over-generalization (RO), where optimal joint actions are undervalued in favor of suboptimal ones. This hinders effective coordination in cooperative tasks, as agents tend to choose actions that are individually rational but collectively suboptimal. To address this issue, we introduce MaxMax Q-Learning (MMQ), which employs an iterative process of sampling and evaluating potential next states, selecting those with maximal Q-values for learning. This approach refines approximations of ideal state transitions, aligning more closely with the optimal joint policy of collaborating agents. We provide theoretical analysis supporting MMQ's potential and present empirical evaluations across various environments susceptible to RO. Our results demonstrate that MMQ frequently outperforms existing baselines, exhibiting enhanced convergence and sample efficiency.

State-Constrained Offline Reinforcement Learning

Traditional offline reinforcement learning methods predominantly operate in a batch-constrained setting. This confines the algorithms to a specific state-action distribution present in the dataset, reducing the effects of distributional shift but restricting the algorithm greatly. In this paper, we alleviate this limitation by introducing a novel framework named \emph{state-constrained} offline reinforcement learning. By exclusively focusing on the dataset's state distribution, our framework significantly enhances learning potential and reduces previous limitations. The proposed setting not only broadens the learning horizon but also improves the ability to combine different trajectories from the dataset effectively, a desirable property inherent in offline reinforcement learning. Our research is underpinned by solid theoretical findings that pave the way for subsequent advancements in this domain. Additionally, we introduce StaCQ, a deep learning algorithm that is both performance-driven on the D4RL benchmark datasets and closely aligned with our theoretical propositions. StaCQ establishes a strong baseline for forthcoming explorations in state-constrained offline reinforcement learning.

REValueD: Regularised Ensemble Value-Decomposition for Factorisable Markov Decision Processes

Discrete-action reinforcement learning algorithms often falter in tasks with high-dimensional discrete action spaces due to the vast number of possible actions. A recent advancement leverages value-decomposition, a concept from multi-agent reinforcement learning, to tackle this challenge. This study delves deep into the effects of this value-decomposition, revealing that whilst it curtails the over-estimation bias inherent to Q-learning algorithms, it amplifies target variance. To counteract this, we present an ensemble of critics to mitigate target variance. Moreover, we introduce a regularisation loss that helps to mitigate the effects that exploratory actions in one dimension can have on the value of optimal actions in other dimensions. Our novel algorithm, REValueD, tested on discretised versions of the DeepMind Control Suite tasks, showcases superior performance, especially in the challenging humanoid and dog tasks. We further dissect the factors influencing REValueD's performance, evaluating the significance of the regularisation loss and the scalability of REValueD with increasing sub-actions per dimension.

Generating a Graph Colouring Heuristic with Deep Q-Learning and Graph Neural Networks

The graph colouring problem consists of assigning labels, or colours, to the vertices of a graph such that no two adjacent vertices share the same colour. In this work we investigate whether deep reinforcement learning can be used to discover a competitive construction heuristic for graph colouring. Our proposed approach, ReLCol, uses deep Q-learning together with a graph neural network for feature extraction, and employs a novel way of parameterising the graph that results in improved performance. Using standard benchmark graphs with varied topologies, we empirically evaluate the benefits and limitations of the heuristic learned by ReLCol relative to existing construction algorithms, and demonstrate that reinforcement learning is a promising direction for further research on the graph colouring problem.

Balancing policy constraint and ensemble size in uncertainty-based offline reinforcement learning

Offline reinforcement learning agents seek optimal policies from fixed data sets. With environmental interaction prohibited, agents face significant challenges in preventing errors in value estimates from compounding and subsequently causing the learning process to collapse. Uncertainty estimation using ensembles compensates for this by penalising high-variance value estimates, allowing agents to learn robust policies based on data-driven actions. However, the requirement for large ensembles to facilitate sufficient penalisation results in significant computational overhead. In this work, we examine the role of policy constraints as a mechanism for regulating uncertainty, and the corresponding balance between level of constraint and ensemble size. By incorporating behavioural cloning into policy updates, we show empirically that sufficient penalisation can be achieved with a much smaller ensemble size, substantially reducing computational demand while retaining state-of-the-art performance on benchmarking tasks. Furthermore, we show how such an approach can facilitate stable online fine tuning, allowing for continued policy improvement while avoiding severe performance drops.

Goal-conditioned Offline Reinforcement Learning through State Space Partitioning

Offline reinforcement learning (RL) aims to infer sequential decision policies using only offline datasets. This is a particularly difficult setup, especially when learning to achieve multiple different goals or outcomes under a given scenario with only sparse rewards. For offline learning of goal-conditioned policies via supervised learning, previous work has shown that an advantage weighted log-likelihood loss guarantees monotonic policy improvement. In this work we argue that, despite its benefits, this approach is still insufficient to fully address the distribution shift and multi-modality problems. The latter is particularly severe in long-horizon tasks where finding a unique and optimal policy that goes from a state to the desired goal is challenging as there may be multiple and potentially conflicting solutions. To tackle these challenges, we propose a complementary advantage-based weighting scheme that introduces an additional source of inductive bias: given a value-based partitioning of the state space, the contribution of actions expected to lead to target regions that are easier to reach, compared to the final goal, is further increased. Empirically, we demonstrate that the proposed approach, Dual-Advantage Weighted Offline Goal-conditioned RL (DAWOG), outperforms several competing offline algorithms in commonly used benchmarks. Analytically, we offer a guarantee that the learnt policy is never worse than the underlying behaviour policy.

Model-based trajectory stitching for improved behavioural cloning and its applications

Behavioural cloning (BC) is a commonly used imitation learning method to infer a sequential decision-making policy from expert demonstrations. However, when the quality of the data is not optimal, the resulting behavioural policy also performs sub-optimally once deployed. Recently, there has been a surge in offline reinforcement learning methods that hold the promise to extract high-quality policies from sub-optimal historical data. A common approach is to perform regularisation during training, encouraging updates during policy evaluation and/or policy improvement to stay close to the underlying data. In this work, we investigate whether an offline approach to improving the quality of the existing data can lead to improved behavioural policies without any changes in the BC algorithm. The proposed data improvement approach - Trajectory Stitching (TS) - generates new trajectories (sequences of states and actions) by `stitching' pairs of states that were disconnected in the original data and generating their connecting new action. By construction, these new transitions are guaranteed to be highly plausible according to probabilistic models of the environment, and to improve a state-value function. We demonstrate that the iterative process of replacing old trajectories with new ones incrementally improves the underlying behavioural policy. Extensive experimental results show that significant performance gains can be achieved using TS over BC policies extracted from the original data. Furthermore, using the D4RL benchmarking suite, we demonstrate that state-of-the-art results are obtained by combining TS with two existing offline learning methodologies reliant on BC, model-based offline planning (MBOP) and policy constraint (TD3+BC).

Improving TD3-BC: Relaxed Policy Constraint for Offline Learning and Stable Online Fine-Tuning

The ability to discover optimal behaviour from fixed data sets has the potential to transfer the successes of reinforcement learning (RL) to domains where data collection is acutely problematic. In this offline setting, a key challenge is overcoming overestimation bias for actions not present in data which, without the ability to correct for via interaction with the environment, can propagate and compound during training, leading to highly sub-optimal policies. One simple method to reduce this bias is to introduce a policy constraint via behavioural cloning (BC), which encourages agents to pick actions closer to the source data. By finding the right balance between RL and BC such approaches have been shown to be surprisingly effective while requiring minimal changes to the underlying algorithms they are based on. To date this balance has been held constant, but in this work we explore the idea of tipping this balance towards RL following initial training. Using TD3-BC, we demonstrate that by continuing to train a policy offline while reducing the influence of the BC component we can produce refined policies that outperform the original baseline, as well as match or exceed the performance of more complex alternatives. Furthermore, we demonstrate such an approach can be used for stable online fine-tuning, allowing policies to be safely improved during deployment.