Search for: All records

While reinforcement learning (RL) has made great advances in scalability, exploration and partial observability are still active research topics. In contrast, Bayesian RL (BRL) provides a principled answer to both state estimation and the exploration-exploitation trade-off, but struggles to scale. To tackle this challenge, BRL frameworks with various prior assumptions have been proposed, with varied success. This work presents a representation-agnostic formulation of BRL under partial observability, unifying the previous models under one theoretical umbrella. To demonstrate its practical significance we also propose a novel derivation, Bayes-Adaptive Deep Dropout rl (BADDr), based on dropout networks. Under this parameterization, in contrast to previous work, the belief over the state and dynamics is a more scalable inference problem. We choose actions through Monte-Carlo tree search and empirically show that our method is competitive with state-of-the-art BRL methods on small domains while being able to solve much larger ones. 

Modern deep reinforcement learning methods have departed from the incremental learning required for eligibility traces, rendering the implementation of the λ-return difficult in this context. In particular, off-policy methods that utilize experience replay remain problematic because their random sampling of minibatches is not conducive to the efficient calculation of λ-returns. Yet replay-based methods are often the most sample efficient, and incorporating λ-returns into them is a viable way to achieve new state-of-the-art performance. Towards this, we propose the first method to enable practical use of λ-returns in arbitrary replay-based methods without relying on other forms of decorrelation such as asynchronous gradient updates. By promoting short sequences of past transitions into a small cache within the replay memory, adjacent λ-returns can be efficiently precomputed by sharing Q-values. Computation is not wasted on experiences that are never sampled, and stored λ-returns behave as stable temporal-difference (TD) targets that replace the target network. Additionally, our method grants the unique ability to observe TD errors prior to sampling; for the first time, transitions can be prioritized by their true significance rather than by a proxy to it. Furthermore, we propose the novel use of the TD error to dynamically select λ-values that facilitate faster learning. We show that these innovations can enhance the performance of DQN when playing Atari 2600 games, even under partial observability. While our work specifically focuses on λ-returns, these ideas are applicable to any multi-step return estimator. 

In real-world multi-robot systems, performing high-quality, collaborative behaviors requires robots to asynchronously reason about high-level action selection at varying time durations. Macro-Action Decentralized Partially Observable Markov Decision Processes (MacDec-POMDPs) provide a general framework for asynchronous decision making under uncertainty in fully cooperative multi-agent tasks. However, multi-agent deep reinforcement learning methods have only been developed for (synchronous) primitive-action problems. This paper proposes two Deep Q-Network (DQN) based methods for learning decentralized and centralized macro-action-value functions with novel macro-action trajectory replay buffers introduced for each case. Evaluations on benchmark problems and a larger domain demonstrate the advantage of learning with macro-actions over primitive-actions and the scalability of our approaches. 

Model-based Bayesian Reinforcement Learning (BRL) provides a principled solution to dealing with the exploration-exploitation trade-off, but such methods typically assume a fully observable environments. The few Bayesian RL methods that are applicable in partially observable domains, such as the Bayes-Adaptive POMDP (BA-POMDP), scale poorly. To address this issue, we introduce the Factored BA-POMDP model (FBA-POMDP), a framework that is able to learn a compact model of the dynamics by exploiting the underlying structure of a POMDP. The FBA-POMDP framework casts the problem as a planning task, for which we adapt the Monte-Carlo Tree Search planning algorithm and develop a belief tracking method to approximate the joint posterior over the state and model variables. Our empirical results show that this method outperforms a number of BRL baselines and is able to learn efficiently when the factorization is known, as well as learn both the factorization and the model parameters simultaneously.