Matthijs T. J. Spaan

Guopeng Li, Moritz A. Zanger, Matthijs T. J. Spaan et al.

Safe robot control requires maximizing return while satisfying safety constraints. In off-policy safe reinforcement learning, reward and safety Q-values are commonly learned by separate critic ensembles, with uncertainty handled independently for each objective. This objective-wise treatment neglects inter-objective correlation and can lead to overly conservative value estimates, thereby reducing sample efficiency. To address this issue, we propose Cholesky-Ordered Projection Q-learning (COP-Q), a safety-first method that incorporates inter-objective covariance into vector-valued Q-value estimation. COP-Q constructs a generalized confidence bound in the joint Q-value space and uses Cholesky factorization to encode objective priority in a sequential form. This preserves conservatism on safety while adaptively reducing excessive conservatism on the reward objective. The resulting estimate is used in both temporal-difference target computation and actor optimization. COP-Q incurs minimal computational overhead and is readily compatible with most existing deep Q-learning frameworks. Experiments on robot locomotion in Brax and safe navigation in Safety-Gymnasium, covering both hard- and soft-safety settings, demonstrate that COP-Q achieves strong safety performance together with competitive or improved sample efficiency relative to representative baselines.

Sebastian Junges, Matthijs T. J. Spaan

Markov decision processes are a ubiquitous formalism for modelling systems with non-deterministic and probabilistic behavior. Verification of these models is subject to the famous state space explosion problem. We alleviate this problem by exploiting a hierarchical structure with repetitive parts. This structure not only occurs naturally in robotics, but also in probabilistic programs describing, e.g., network protocols. Such programs often repeatedly call a subroutine with similar behavior. In this paper, we focus on a local case, in which the subroutines have a limited effect on the overall system state. The key ideas to accelerate analysis of such programs are (1) to treat the behavior of the subroutine as uncertain and only remove this uncertainty by a detailed analysis if needed, and (2) to abstract similar subroutines into a parametric template, and then analyse this template. These two ideas are embedded into an abstraction-refinement loop that analyses hierarchical MDPs. A prototypical implementation shows the efficacy of the approach.

Maris F. L. Galesloot, Marnix Suilen, Thiago D. Simão et al.

Robust POMDPs extend classical POMDPs to incorporate model uncertainty using so-called uncertainty sets on the transition and observation functions, effectively defining ranges of probabilities. Policies for robust POMDPs must be (1) memory-based to account for partial observability and (2) robust against model uncertainty to account for the worst-case probability instances from the uncertainty sets. To compute such robust memory-based policies, we propose the pessimistic iterative planning (PIP) framework, which alternates between (1) selecting pessimistic POMDPs via worst-case probability instances from the uncertainty sets, and (2) computing finite-state controllers (FSCs) for these pessimistic POMDPs. Within PIP, we propose the rFSCNet algorithm, which optimizes a recurrent neural network to compute the FSCs. The empirical evaluation shows that rFSCNet can compute better-performing robust policies than several baselines and a state-of-the-art robust POMDP solver.

Qisong Yang, Thiago D. Simão, Nils Jansen et al.

Safety is critical to broadening the application of reinforcement learning (RL). Often, we train RL agents in a controlled environment, such as a laboratory, before deploying them in the real world. However, the real-world target task might be unknown prior to deployment. Reward-free RL trains an agent without the reward to adapt quickly once the reward is revealed. We consider the constrained reward-free setting, where an agent (the guide) learns to explore safely without the reward signal. This agent is trained in a controlled environment, which allows unsafe interactions and still provides the safety signal. After the target task is revealed, safety violations are not allowed anymore. Thus, the guide is leveraged to compose a safe behaviour policy. Drawing from transfer learning, we also regularize a target policy (the student) towards the guide while the student is unreliable and gradually eliminate the influence of the guide as training progresses. The empirical analysis shows that this method can achieve safe transfer learning and helps the student solve the target task faster.

Miguel Suau, Matthijs T. J. Spaan, Frans A. Oliehoek

Reinforcement learning agents tend to develop habits that are effective only under specific policies. Following an initial exploration phase where agents try out different actions, they eventually converge onto a particular policy. As this occurs, the distribution over state-action trajectories becomes narrower, leading agents to repeatedly experience the same transitions. This repetitive exposure fosters spurious correlations between certain observations and rewards. Agents may then pick up on these correlations and develop simplistic habits tailored to the specific set of trajectories dictated by their policy. The problem is that these habits may yield incorrect outcomes when agents are forced to deviate from their typical trajectories, prompted by changes in the environment. This paper presents a mathematical characterization of this phenomenon, termed policy confounding, and illustrates, through a series of examples, the circumstances under which it occurs.

Moritz A. Zanger, Wendelin Böhmer, Matthijs T. J. Spaan

In contrast to classical reinforcement learning (RL), distributional RL algorithms aim to learn the distribution of returns rather than their expected value. Since the nature of the return distribution is generally unknown a priori or arbitrarily complex, a common approach finds approximations within a set of representable, parametric distributions. Typically, this involves a projection of the unconstrained distribution onto the set of simplified distributions. We argue that this projection step entails a strong inductive bias when coupled with neural networks and gradient descent, thereby profoundly impacting the generalization behavior of learned models. In order to facilitate reliable uncertainty estimation through diversity, we study the combination of several different projections and representations in a distributional ensemble. We establish theoretical properties of such projection ensembles and derive an algorithm that uses ensemble disagreement, measured by the average 1-Wasserstein distance, as a bonus for deep exploration. We evaluate our algorithm on the behavior suite benchmark and VizDoom and find that diverse projection ensembles lead to significant performance improvements over existing methods on a variety of tasks with the most pronounced gains in directed exploration problems.

Yaniv Oren, Villiam Vadocz, Matthijs T. J. Spaan et al.

The AlphaZero/MuZero (A/MZ) family of algorithms has achieved remarkable success across various challenging domains by integrating Monte Carlo Tree Search (MCTS) with learned models. Learned models introduce epistemic uncertainty, which is caused by learning from limited data and is useful for exploration in sparse reward environments. MCTS does not account for the propagation of this uncertainty however. To address this, we introduce Epistemic MCTS (EMCTS): a theoretically motivated approach to account for the epistemic uncertainty in search and harness the search for deep exploration. In the challenging sparse-reward task of writing code in the Assembly language {\sc subleq}, AZ paired with our method achieves significantly higher sample efficiency over baseline AZ. Search with EMCTS solves variations of the commonly used hard-exploration benchmark Deep Sea - which baseline A/MZ are practically unable to solve - much faster than an otherwise equivalent method that does not use search for uncertainty estimation, demonstrating significant benefits from search for epistemic uncertainty estimation.

Max Weltevrede, Matthijs T. J. Spaan, Wendelin Böhmer

In reinforcement learning (RL), key components of many algorithms are the exploration strategy and replay buffer. These strategies regulate what environment data is collected and trained on and have been extensively studied in the RL literature. In this paper, we investigate the impact of these components in the context of generalisation in multi-task RL. We investigate the hypothesis that collecting and training on more diverse data from the training environments will improve zero-shot generalisation to new tasks. We motivate mathematically and show empirically that generalisation to tasks that are "reachable'' during training is improved by increasing the diversity of transitions in the replay buffer. Furthermore, we show empirically that this same strategy also shows improvement for generalisation to similar but "unreachable'' tasks which could be due to improved generalisation of the learned latent representations.

Miguel Suau, Jinke He, Mustafa Mert Çelikok et al.

Due to its high sample complexity, simulation is, as of today, critical for the successful application of reinforcement learning. Many real-world problems, however, exhibit overly complex dynamics, which makes their full-scale simulation computationally slow. In this paper, we show how to decompose large networked systems of many agents into multiple local components such that we can build separate simulators that run independently and in parallel. To monitor the influence that the different local components exert on one another, each of these simulators is equipped with a learned model that is periodically trained on real trajectories. Our empirical results reveal that distributing the simulation among different processes not only makes it possible to train large multi-agent systems in just a few hours but also helps mitigate the negative effects of simultaneous learning.

Guopeng Li, Matthijs T. J. Spaan, Julian F. P. Kooij

When safety is formulated as a limit of cumulative cost, safe reinforcement learning (RL) aims to learn policies that maximize return subject to the cost constraint in data collection and deployment. Off-policy safe RL methods, although offering high sample efficiency, suffer from constraint violations due to cost-agnostic exploration and estimation bias in cumulative cost. To address this issue, we propose Constrained Optimistic eXploration Q-learning (COX-Q), an off-policy safe RL algorithm that integrates cost-bounded online exploration and conservative offline distributional value learning. First, we introduce a novel cost-constrained optimistic exploration strategy that resolves gradient conflicts between reward and cost in the action space and adaptively adjusts the trust region to control the training cost. Second, we adopt truncated quantile critics to stabilize the cost value learning. Quantile critics also quantify epistemic uncertainty to guide exploration. Experiments on safe velocity, safe navigation, and autonomous driving tasks demonstrate that COX-Q achieves high sample efficiency, competitive test safety performance, and controlled data collection cost. The results highlight COX-Q as a promising RL method for safety-critical applications.

Moritz A. Zanger, Yijun Wu, Pascal R. Van der Vaart et al.

Uncertainty quantification is central to safe and efficient deployments of deep learning models, yet many computationally practical methods lack lacking rigorous theoretical motivation. Random network distillation (RND) is a lightweight technique that measures novelty via prediction errors against a fixed random target. While empirically effective, it has remained unclear what uncertainties RND measures and how its estimates relate to other approaches, e.g. Bayesian inference or deep ensembles. This paper establishes these missing theoretical connections by analyzing RND within the neural tangent kernel framework in the limit of infinite network width. Our analysis reveals two central findings in this limit: (1) The uncertainty signal from RND -- its squared self-predictive error -- is equivalent to the predictive variance of a deep ensemble. (2) By constructing a specific RND target function, we show that the RND error distribution can be made to mirror the centered posterior predictive distribution of Bayesian inference with wide neural networks. Based on this equivalence, we moreover devise a posterior sampling algorithm that generates i.i.d. samples from an exact Bayesian posterior predictive distribution using this modified \textit{Bayesian RND} model. Collectively, our findings provide a unified theoretical perspective that places RND within the principled frameworks of deep ensembles and Bayesian inference, and offer new avenues for efficient yet theoretically grounded uncertainty quantification methods.

Caroline Horsch, Laurens Engwegen, Max Weltevrede et al.

In reinforcement learning, abstraction methods that remove unnecessary information from the observation are commonly used to learn policies which generalize better to unseen tasks. However, these methods often overlook a crucial weakness: the function which extracts the reduced-information representation has unknown generalization ability in unseen observations. In this paper, we address this problem by presenting an information removal method which more reliably generalizes to new states. We accomplish this by using a learned masking function which operates on, and is integrated with, the attention weights within an attention-based policy network. We demonstrate that our method significantly improves policy generalization to unseen tasks in the Procgen benchmark compared to standard PPO and masking approaches.

Philip Lippmann, Matthijs T. J. Spaan, Jie Yang

Intelligent agents designed for interactive environments face significant challenges in text-based games, a domain that demands complex reasoning and adaptability. While agents based on large language models (LLMs) using self-reflection have shown promise, they struggle when initially successful and exhibit reduced effectiveness when using smaller LLMs. We introduce Sweet&Sour, a novel approach that addresses these limitations in existing reflection methods by incorporating positive experiences and managed memory to enrich the context available to the agent at decision time. Our comprehensive analysis spans both closed- and open-source LLMs and demonstrates the effectiveness of Sweet&Sour in improving agent performance, particularly in scenarios where previous approaches fall short.

Yaniv Oren, Viliam Vadocz, Joery A. de Vries et al.

Monte Carlo Tree Search (MCTS) is a widely used approach for policy improvement through search with increasing popularity for real world applications. Due to the sequential and deterministic nature of its search, runtime-scaling of MCTS with parallel compute remains a major challenge. We introduce Particle MCTS (PMCTS), to our knowledge the first principled parallel MCTS algorithm which is suited for neural network evaluations and can preserve formal policy improvement guarantees. Empirically, PMCTS scales well with parallel compute and significantly outperforms the popular heuristic-based baselines across domains.

Moritz A. Zanger, Pascal R. Van der Vaart, Wendelin Böhmer et al.

Uncertainty quantification is a critical aspect of reinforcement learning and deep learning, with numerous applications ranging from efficient exploration and stable offline reinforcement learning to outlier detection in medical diagnostics. The scale of modern neural networks, however, complicates the use of many theoretically well-motivated approaches such as full Bayesian inference. Approximate methods like deep ensembles can provide reliable uncertainty estimates but still remain computationally expensive. In this work, we propose contextual similarity distillation, a novel approach that explicitly estimates the variance of an ensemble of deep neural networks with a single model, without ever learning or evaluating such an ensemble in the first place. Our method builds on the predictable learning dynamics of wide neural networks, governed by the neural tangent kernel, to derive an efficient approximation of the predictive variance of an infinite ensemble. Specifically, we reinterpret the computation of ensemble variance as a supervised regression problem with kernel similarities as regression targets. The resulting model can estimate predictive variance at inference time with a single forward pass, and can make use of unlabeled target-domain data or data augmentations to refine its uncertainty estimates. We empirically validate our method across a variety of out-of-distribution detection benchmarks and sparse-reward reinforcement learning environments. We find that our single-model method performs competitively and sometimes superior to ensemble-based baselines and serves as a reliable signal for efficient exploration. These results, we believe, position contextual similarity distillation as a principled and scalable alternative for uncertainty quantification in reinforcement learning and general deep learning.

Davide Mambelli, Stephan Bongers, Onno Zoeter et al.

Policy gradient methods are widely adopted reinforcement learning algorithms for tasks with continuous action spaces. These methods succeeded in many application domains, however, because of their notorious sample inefficiency their use remains limited to problems where fast and accurate simulations are available. A common way to improve sample efficiency is to modify their objective function to be computable from off-policy samples without importance sampling. A well-established off-policy objective is the excursion objective. This work studies the difference between the excursion objective and the traditional on-policy objective, which we refer to as the on-off gap. We provide the first theoretical analysis showing conditions to reduce the on-off gap while establishing empirical evidence of shortfalls arising when these conditions are not met.

Philip Lippmann, Matthijs T. J. Spaan, Jie Yang

Language models (LMs) have achieved impressive accuracy across a variety of tasks but remain vulnerable to high-confidence misclassifications, also referred to as unknown unknowns (UUs). These UUs cluster into blind spots in the feature space, leading to significant risks in high-stakes applications. This is particularly relevant for smaller, lightweight LMs that are more susceptible to such errors. While the identification of UUs has been extensively studied, their mitigation remains an open challenge, including how to use identified UUs to eliminate unseen blind spots. In this work, we propose a novel approach to address blind spot mitigation through the use of intelligent agents -- either humans or large LMs -- as teachers to characterize UU-type errors. By leveraging the generalization capabilities of intelligent agents, we identify patterns in high-confidence misclassifications and use them to generate targeted synthetic samples to improve model robustness and reduce blind spots. We conduct an extensive evaluation of our method on three classification tasks and demonstrate its effectiveness in reducing the number of UUs, all while maintaining a similar level of accuracy. We find that the effectiveness of human computation has a high ceiling but is highly dependent on familiarity with the underlying task. Moreover, the cost gap between humans and LMs surpasses an order of magnitude, as LMs attain human-like generalization and generation performance while being more scalable.

Joery A. de Vries, Jinke He, Yaniv Oren et al.

Optimally trading-off exploration and exploitation is the holy grail of reinforcement learning as it promises maximal data-efficiency for solving any task. Bayes-optimal agents achieve this, but obtaining the belief-state and performing planning are both typically intractable. Although deep learning methods can greatly help in scaling this computation, existing methods are still costly to train. To accelerate this, this paper proposes a variational framework for learning and planning in Bayes-adaptive Markov decision processes that coalesces variational belief learning, sequential Monte-Carlo planning, and meta-reinforcement learning. In a single-GPU setup, our new method VariBASeD exhibits favorable scaling to larger planning budgets, improving sample- and runtime-efficiency over prior methods.

Yaniv Oren, Joery A. de Vries, Pascal R. van der Vaart et al.

Model-based reinforcement learning (RL) methods that leverage search are responsible for many milestone breakthroughs in RL. Sequential Monte Carlo (SMC) recently emerged as an alternative to the Monte Carlo Tree Search (MCTS) algorithm which drove these breakthroughs. SMC is easier to parallelize and more suitable to GPU acceleration. However, it also suffers from large variance and path degeneracy which prevent it from scaling well with increased search depth, i.e., increased sequential compute. To address these problems, we introduce Twice Sequential Monte Carlo Tree Search (TSMCTS). Across discrete and continuous environments TSMCTS outperforms the SMC baseline as well as a popular modern version of MCTS. Through variance reduction and mitigation of path degeneracy, TSMCTS scales favorably with sequential compute while retaining the properties that make SMC natural to parallelize.

Pascal R. van der Vaart, Neil Yorke-Smith, Matthijs T. J. Spaan

Uncertainty quantification in reinforcement learning can greatly improve exploration and robustness. Approximate Bayesian approaches have recently been popularized to quantify uncertainty in model-free algorithms. However, so far the focus has been on improving the accuracy of the posterior approximation, instead of studying the accuracy of the prior and likelihood assumptions underlying the posterior. In this work, we demonstrate that there is a cold posterior effect in Bayesian deep Q-learning, where contrary to theory, performance increases when reducing the temperature of the posterior. To identify and overcome likely causes, we challenge common assumptions made on the likelihood and priors in Bayesian model-free algorithms. We empirically study prior distributions and show through statistical tests that the common Gaussian likelihood assumption is frequently violated. We argue that developing more suitable likelihoods and priors should be a key focus in future Bayesian reinforcement learning research and we offer simple, implementable solutions for better priors in deep Q-learning that lead to more performant Bayesian algorithms.

Moritz A. Zanger, Max Weltevrede, Yaniv Oren et al.

Estimating epistemic uncertainty in value functions is a crucial challenge for many aspects of reinforcement learning (RL), including efficient exploration, safe decision-making, and offline RL. While deep ensembles provide a robust method for quantifying value uncertainty, they come with significant computational overhead. Single-model methods, while computationally favorable, often rely on heuristics and typically require additional propagation mechanisms for myopic uncertainty estimates. In this work we introduce universal value-function uncertainties (UVU), which, similar in spirit to random network distillation (RND), quantify uncertainty as squared prediction errors between an online learner and a fixed, randomly initialized target network. Unlike RND, UVU errors reflect policy-conditional value uncertainty, incorporating the future uncertainties any given policy may encounter. This is due to the training procedure employed in UVU: the online network is trained using temporal difference learning with a synthetic reward derived from the fixed, randomly initialized target network. We provide an extensive theoretical analysis of our approach using neural tangent kernel (NTK) theory and show that in the limit of infinite network width, UVU errors are exactly equivalent to the variance of an ensemble of independent universal value functions. Empirically, we show that UVU achieves equal performance to large ensembles on challenging multi-task offline RL settings, while offering simplicity and substantial computational savings.

Max Weltevrede, Moritz A. Zanger, Matthijs T. J. Spaan et al.

In the zero-shot policy transfer setting in reinforcement learning, the goal is to train an agent on a fixed set of training environments so that it can generalise to similar, but unseen, testing environments. Previous work has shown that policy distillation after training can sometimes produce a policy that outperforms the original in the testing environments. However, it is not yet entirely clear why that is, or what data should be used to distil the policy. In this paper, we prove, under certain assumptions, a generalisation bound for policy distillation after training. The theory provides two practical insights: for improved generalisation, you should 1) train an ensemble of distilled policies, and 2) distil it on as much data from the training environments as possible. We empirically verify that these insights hold in more general settings, when the assumptions required for the theory no longer hold. Finally, we demonstrate that an ensemble of policies distilled on a diverse dataset can generalise significantly better than the original agent.

Joery A. de Vries, Jinke He, Mathijs M. de Weerdt et al.

Meta-reinforcement learning trains a single reinforcement learning agent on a distribution of tasks to quickly generalize to new tasks outside of the training set at test time. From a Bayesian perspective, one can interpret this as performing amortized variational inference on the posterior distribution over training tasks. Among the various meta-reinforcement learning approaches, a common method is to represent this distribution with a point-estimate using a recurrent neural network. We show how one can augment this point estimate to give full distributions through the Laplace approximation, either at the start of, during, or after learning, without modifying the base model architecture. With our approximation, we are able to estimate distribution statistics (e.g., the entropy) of non-Bayesian agents and observe that point-estimate based methods produce overconfident estimators while not satisfying consistency. Furthermore, when comparing our approach to full-distribution based learning of the task posterior, our method performs on par with variational baselines while having much fewer parameters.

Sterre Lutz, Matthijs T. J. Spaan, Anna Lukina

Autonomous systems operating in the real world encounter a range of uncertainties. Probabilistic neural Lyapunov certification is a powerful approach to proving safety of nonlinear stochastic dynamical systems. When faced with changes beyond the modeled uncertainties, e.g., unidentified obstacles, probabilistic certificates must be transferred to the new system dynamics. However, even when the changes are localized in a known part of the state space, state-of-the-art requires complete re-certification, which is particularly costly for neural certificates. We introduce VeRecycle, the first framework to formally reclaim guarantees for discrete-time stochastic dynamical systems. VeRecycle efficiently reuses probabilistic certificates when the system dynamics deviate only in a given subset of states. We present a general theoretical justification and algorithmic implementation. Our experimental evaluation shows scenarios where VeRecycle both saves significant computational effort and achieves competitive probabilistic guarantees in compositional neural control.

Joery A. de Vries, Jinke He, Yaniv Oren et al.

Monte-Carlo tree search (MCTS) has driven many recent breakthroughs in deep reinforcement learning (RL). However, scaling MCTS to parallel compute has proven challenging in practice which has motivated alternative planners like sequential Monte-Carlo (SMC). Many of these SMC methods adopt particle filters for smoothing through a reformulation of RL as a policy inference problem. Yet, persisting design choices of these particle filters often conflict with the aim of online planning in RL, which is to obtain a policy improvement at the start of planning. Drawing inspiration from MCTS, we tailor SMC planners specifically for RL by improving data generation within the planner through constrained action sampling and explicit terminal state handling, as well as improving policy and value target estimation. This leads to our Trust-Region Twisted SMC (TRT-SMC), which shows improved runtime and sample-efficiency over baseline MCTS and SMC methods in both discrete and continuous domains.

Max Weltevrede, Felix Kaubek, Matthijs T. J. Spaan et al.

One of the remaining challenges in reinforcement learning is to develop agents that can generalise to novel scenarios they might encounter once deployed. This challenge is often framed in a multi-task setting where agents train on a fixed set of tasks and have to generalise to new tasks. Recent work has shown that in this setting increased exploration during training can be leveraged to increase the generalisation performance of the agent. This makes sense when the states encountered during testing can actually be explored during training. In this paper, we provide intuition why exploration can also benefit generalisation to states that cannot be explicitly encountered during training. Additionally, we propose a novel method Explore-Go that exploits this intuition by increasing the number of states on which the agent trains. Explore-Go effectively increases the starting state distribution of the agent and as a result can be used in conjunction with most existing on-policy or off-policy reinforcement learning algorithms. We show empirically that our method can increase generalisation performance in an illustrative environment and on the Procgen benchmark.

Yaniv Oren, Moritz A. Zanger, Pascal R. van der Vaart et al.

To learn approximately optimal acting policies for decision problems, modern Actor Critic algorithms rely on deep Neural Networks (DNNs) to parameterize the acting policy and greedification operators to iteratively improve it. The reliance on DNNs suggests an improvement that is gradient based, which is per step much less greedy than the improvement possible by greedier operators such as the greedy update used by Q-learning algorithms. On the other hand, slow changes to the policy can also be beneficial for the stability of the learning process, resulting in a tradeoff between greedification and stability. To better address this tradeoff, we propose to decouple the acting policy from the policy evaluated by the critic. This allows the agent to separately improve the critic's policy (e.g. value improvement) with greedier updates while maintaining the slow gradient-based improvement to the parameterized acting policy. We investigate the convergence of this approach using the popular analysis scheme of generalized Policy Iteration in the finite-horizon domain. Empirically, incorporating value-improvement into the popular off-policy actor-critic algorithms TD3 and SAC significantly improves or matches performance over their respective baselines, across different environments from the DeepMind continuous control domain, with negligible compute and implementation cost.

Miguel Suau, Jinke He, Matthijs T. J. Spaan et al.

Learning effective policies for real-world problems is still an open challenge for the field of reinforcement learning (RL). The main limitation being the amount of data needed and the pace at which that data can be obtained. In this paper, we study how to build lightweight simulators of complicated systems that can run sufficiently fast for deep RL to be applicable. We focus on domains where agents interact with a reduced portion of a larger environment while still being affected by the global dynamics. Our method combines the use of local simulators with learned models that mimic the influence of the global system. The experiments reveal that incorporating this idea into the deep RL workflow can considerably accelerate the training process and presents several opportunities for the future.

Yash Satsangi, Shimon Whiteson, Frans A. Oliehoek et al.

In active perception tasks, an agent aims to select sensory actions that reduce its uncertainty about one or more hidden variables. While partially observable Markov decision processes (POMDPs) provide a natural model for such problems, reward functions that directly penalize uncertainty in the agent's belief can remove the piecewise-linear and convex property of the value function required by most POMDP planners. Furthermore, as the number of sensors available to the agent grows, the computational cost of POMDP planning grows exponentially with it, making POMDP planning infeasible with traditional methods. In this article, we address a twofold challenge of modeling and planning for active perception tasks. We show the mathematical equivalence of $ρ$POMDP and POMDP-IR, two frameworks for modeling active perception tasks, that restore the PWLC property of the value function. To efficiently plan for active perception tasks, we identify and exploit the independence properties of POMDP-IR to reduce the computational cost of solving POMDP-IR (and $ρ$POMDP). We propose greedy point-based value iteration (PBVI), a new POMDP planning method that uses greedy maximization to greatly improve scalability in the action space of an active perception POMDP. Furthermore, we show that, under certain conditions, including submodularity, the value function computed using greedy PBVI is guaranteed to have bounded error with respect to the optimal value function. We establish the conditions under which the value function of an active perception POMDP is guaranteed to be submodular. Finally, we present a detailed empirical analysis on a dataset collected from a multi-camera tracking system employed in a shopping mall. Our method achieves similar performance to existing methods but at a fraction of the computational cost leading to better scalability for solving active perception tasks.

Joris Scharpff, Diederik M. Roijers, Frans A. Oliehoek et al.

In cooperative multi-agent sequential decision making under uncertainty, agents must coordinate to find an optimal joint policy that maximises joint value. Typical algorithms exploit additive structure in the value function, but in the fully-observable multi-agent MDP setting (MMDP) such structure is not present. We propose a new optimal solver for transition-independent MMDPs, in which agents can only affect their own state but their reward depends on joint transitions. We represent these dependencies compactly in conditional return graphs (CRGs). Using CRGs the value of a joint policy and the bounds on partially specified joint policies can be efficiently computed. We propose CoRe, a novel branch-and-bound policy search algorithm building on CRGs. CoRe typically requires less runtime than the available alternatives and finds solutions to problems previously unsolvable.

Frans A. Oliehoek, Matthijs T. J. Spaan, Stefan Witwicki

Recent years have seen the development of methods for multiagent planning under uncertainty that scale to tens or even hundreds of agents. However, most of these methods either make restrictive assumptions on the problem domain, or provide approximate solutions without any guarantees on quality. Methods in the former category typically build on heuristic search using upper bounds on the value function. Unfortunately, no techniques exist to compute such upper bounds for problems with non-factored value functions. To allow for meaningful benchmarking through measurable quality guarantees on a very general class of problems, this paper introduces a family of influence-optimistic upper bounds for factored decentralized partially observable Markov decision processes (Dec-POMDPs) that do not have factored value functions. Intuitively, we derive bounds on very large multiagent planning problems by subdividing them in sub-problems, and at each of these sub-problems making optimistic assumptions with respect to the influence that will be exerted by the rest of the system. We numerically compare the different upper bounds and demonstrate how we can achieve a non-trivial guarantee that a heuristic solution for problems with hundreds of agents is close to optimal. Furthermore, we provide evidence that the upper bounds may improve the effectiveness of heuristic influence search, and discuss further potential applications to multiagent planning.

Frans Adriaan Oliehoek, Matthijs T. J. Spaan, Christopher Amato et al.

This article presents the state-of-the-art in optimal solution methods for decentralized partially observable Markov decision processes (Dec-POMDPs), which are general models for collaborative multiagent planning under uncertainty. Building off the generalized multiagent A* (GMAA*) algorithm, which reduces the problem to a tree of one-shot collaborative Bayesian games (CBGs), we describe several advances that greatly expand the range of Dec-POMDPs that can be solved optimally. First, we introduce lossless incremental clustering of the CBGs solved by GMAA*, which achieves exponential speedups without sacrificing optimality. Second, we introduce incremental expansion of nodes in the GMAA* search tree, which avoids the need to expand all children, the number of which is in the worst case doubly exponential in the nodes depth. This is particularly beneficial when little clustering is possible. In addition, we introduce new hybrid heuristic representations that are more compact and thereby enable the solution of larger Dec-POMDPs. We provide theoretical guarantees that, when a suitable heuristic is used, both incremental clustering and incremental expansion yield algorithms that are both complete and search equivalent. Finally, we present extensive empirical results demonstrating that GMAA*-ICE, an algorithm that synthesizes these advances, can optimally solve Dec-POMDPs of unprecedented size.

Frans A. Oliehoek, Shimon Whiteson, Matthijs T. J. Spaan

Cooperative Bayesian games (BGs) can model decision-making problems for teams of agents under imperfect information, but require space and computation time that is exponential in the number of agents. While agent independence has been used to mitigate these problems in perfect information settings, we propose a novel approach for BGs based on the observation that BGs additionally possess a different types of structure, which we call type independence. We propose a factor graph representation that captures both forms of independence and present a theoretical analysis showing that non-serial dynamic programming cannot effectively exploit type independence, while Max-Sum can. Experimental results demonstrate that our approach can tackle cooperative Bayesian games of unprecedented size.