Some of the most powerful reinforcement learning frameworks use planning for action selection. Interestingly, their planning horizon is either fixed or determined arbitrarily by the state visitation history. Here, we expand beyond the naive fixed horizon and propose a theoretically justified strategy for adaptive selection of the planning horizon as a function of the state-dependent value estimate. We propose two variants for lookahead selection and analyze the trade-off between iteration count and computational complexity per iteration. We then devise a corresponding deep Q-network algorithm with an adaptive tree search horizon. We separate the value estimation per depth to compensate for the off-policy discrepancy between depths. Lastly, we demonstrate the efficacy of our adaptive lookahead method in a maze environment and Atari.
Real-time heuristic search is a popular model of acting and learning in intelligent autonomous agents. Learning real-time search agents improve their performance over time by acquiring and refining a value function guiding the application of their actions. As computing the perfect value function is typically intractable, a heuristic approximation is acquired instead. Most studies of learning in real-time search (and reinforcement learning) assume that a simple value-function-greedy policy is used to select actions. This is in contrast to practice, where high-performance is usually attained by …
National Tsing Hua University（国立清华大学）；TU Darmstadt（达姆施塔特工业大学）；Max Planck Institute for Intelligent Systems（马克斯·普朗克智能系统研究所）
Model-based Reinforcement Learning (MBRL) allows data-efficient learning which is required in real world applications such as robotics. However, despite the impressive data-efficiency, MBRL does not achieve the final performance of state-of-the-art Model-free Reinforcement Learning (MFRL) methods. We leverage the strengths of both realms and propose an approach that obtains high performance with a small amount of data. In particular, we combine MFRL and Model Predictive Control (MPC). While MFRL's strength in exploration allows us to train a better forward dynamics model for MPC, MPC improves …
Robotics Institute（机器人研究所）；Carnegie Mellon University（卡内基梅隆大学）；The University of Texas at Austin（得克萨斯大学奥斯汀分校）
Reinforcement learning (RL) in low-data and risk-sensitive domains requires performant and flexible deployment policies that can readily incorporate constraints during deployment. One such class of policies are the semi-parametric H-step lookahead policies, which select actions using trajectory optimization over a dynamics model for a fixed horizon with a terminal value function. In this work, we investigate a novel instantiation of H-step lookahead with a learned model and a terminal value function learned by a model-free off-policy algorithm, named Learning Off-Policy with Online Planning (L…
Technion, Israel（以色列理工学院）；INRIA, Villers les Nancy, France（法国南希-比莱尔斯法国国家信息与自动化研究所）
Finite-horizon lookahead policies are abundantly used in Reinforcement Learning and demonstrate impressive empirical success. Usually, the lookahead policies are implemented with specific planning methods such as Monte Carlo Tree Search (e.g. in AlphaZero). Referring to the planning problem as tree search, a reasonable practice in these implementations is to back up the value only at the leaves while the information obtained at the root is not leveraged other than for updating the policy. Here, we question the potency of this approach. Namely, the latter procedure is non-contractive in general…
This paper proposes a novel deep reinforcement learning (RL) architecture, called Value Prediction Network (VPN), which integrates model-free and model-based RL methods into a single neural network. In contrast to typical model-based RL methods, VPN learns a dynamics model whose abstract states are trained to make option-conditional predictions of future values (discounted sum of rewards) rather than of future observations. Our experimental results show that VPN has several advantages over both model-free and model-based baselines in a stochastic environment where careful planning is required …
University of Warsaw（华沙大学）；Institute of Mathematics of the Polish Academy of Sciences（波兰科学院数学研究所）
Recent breakthroughs in Artificial Intelligence have shown that the combination of tree-based planning with deep learning can lead to superior performance. We present Adaptive Entropy Tree Search (ANTS) - a novel algorithm combining planning and learning in the maximum entropy paradigm. Through a comprehensive suite of experiments on the Atari benchmark we show that ANTS significantly outperforms PUCT, the planning component of the state-of-the-art AlphaZero system. ANTS builds upon recent work on maximum entropy planning methods - which however, as we show, fail in combination with learning. …
Planning under model uncertainty is a fundamental problem across many applications of decision making and learning. In this paper, we propose the Robust Adaptive Monte Carlo Planning (RAMCP) algorithm, which allows computation of risk-sensitive Bayes-adaptive policies that optimally trade off exploration, exploitation, and robustness. RAMCP formulates the risk-sensitive planning problem as a two-player zero-sum game, in which an adversary perturbs the agent's belief over the models. We introduce two versions of the RAMCP algorithm. The first, RAMCP-F, converges to an optimal risk-sensitive pol…
Facebook AI Research（脸书人工智能研究院）；UC Berkeley（加州大学伯克利分校）
Lookahead search has been a critical component of recent AI successes, such as in the games of chess, go, and poker. However, the search methods used in these games, and in many other settings, are tabular. Tabular search methods do not scale well with the size of the search space, and this problem is exacerbated by stochasticity and partial observability. In this work we replace tabular search with online model-based fine-tuning of a policy neural network via reinforcement learning, and show that this approach outperforms state-of-the-art search algorithms in benchmark settings. In particular…
Partially Observable Markov Decision Processes (POMDPs) provide a rich framework for sequential decision-making under uncertainty in stochastic domains. However, solving a POMDP is often intractable except for small problems due to their complexity. Here, we focus on online approaches that alleviate the computational complexity by computing good local policies at each decision step during the execution. Online algorithms generally consist of a lookahead search to find the best action to execute at each time step in an environment. Our objectives here are to survey the various existing online P…
The computational costs of inference and planning have confined Bayesian model-based reinforcement learning to one of two dismal fates: powerful Bayes-adaptive planning but only for simplistic models, or powerful, Bayesian non-parametric models but using simple, myopic planning strategies such as Thompson sampling. We ask whether it is feasible and truly beneficial to combine rich probabilistic models with a closer approximation to fully Bayesian planning. First, we use a collection of counterexamples to show formal problems with the over-optimism inherent in Thompson sampling. Then we leverag…
Learning to Plan, Planning to Learn: Adaptive Hierarchical RL-MPC for Sample-Efficient Decision Making
Toyota Research Institute（丰田研究院）；Learning to Plan, Planning to Learn: Adaptive Hierarchical RL-MPC for Sample-Efficient Decision Making（学习规划，规划学习：自适应分层RL-MPC用于样本高效决策）
Facebook AI Research（脸书人工智能研究院）；Microsoft Research（微软研究院）；Technion（以色列理工学院）；Google Research（谷歌研究院）；Nvidia Research（英伟达研究院）
Real Time Dynamic Programming (RTDP) is an online algorithm based on Dynamic Programming (DP) that acts by 1-step greedy planning. Unlike DP, RTDP does not require access to the entire state space, i.e., it explicitly handles the exploration. This fact makes RTDP particularly appealing when the state space is large and it is not possible to update all states simultaneously. In this we devise a multi-step greedy RTDP algorithm, which we call $h$-RTDP, that replaces the 1-step greedy policy with a $h$-step lookahead policy. We analyze $h$-RTDP in its exact form and establish that increasing the …
We consider some classical optimization problems in path planning and network transport, and we introduce new auction-based algorithms for their optimal and suboptimal solution. The algorithms are based on mathematical ideas that are related to competitive bidding by persons for objects and the attendant market equilibrium, which underlie auction processes. However, the starting point of our algorithms is different, namely weighted and unweighted path construction in directed graphs, rather than assignment of persons to objects. The new algorithms have several potential advantages over existin…
Graph Sparse Sampling: Breaking the Curse of the Horizon in Continuous MDP Planning
Technion Autonomous Systems Program (TASP), Technion – Israel Institute of Technology（以色列理工学院自主系统计划（TASP），以色列理工学院）；Stephen B. Klein Faculty of Aerospace Engineering, Technion – Israel Institute of Technology（以色列理工学院斯蒂芬·B·克莱因航空航天工程学院）；Faculty of Data and Decision Sciences, Technion – Israel Institute of Technology（以色列理工学院数据与决策科学学院）
Partially Observable Markov Decision Processes (POMDP) is a widely used model to represent the interaction of an environment and an agent, under state uncertainty. Since the agent does not observe the environment state, its uncertainty is typically represented through a probabilistic belief. While the set of possible beliefs is infinite, making exact planning intractable, the belief space's complexity (and hence planning complexity) is characterized by its covering number. Many POMDP solvers uniformly discretize the belief space and give the planning error in terms of the (typically unknown) c…
Adaptive Informative Path Planning (AIPP) problems model an agent tasked with obtaining information subject to resource constraints in unknown, partially observable environments. Existing work on AIPP has focused on representing observations about the world as a result of agent movement. We formulate the more general setting where the agent may choose between different sensors at the cost of some energy, in addition to traversing the environment to gather information. We call this problem AIPPMS (MS for Multimodal Sensing). AIPPMS requires reasoning jointly about the effects of sensing and mov…
Stanford University（斯坦福大学）；NASA Ames Research Center（美国国家航空航天局艾姆斯研究中心）
Adaptive Informative Path Planning with Multimodal Sensing (AIPPMS) considers the problem of an agent equipped with multiple sensors, each with different sensing accuracy and energy costs. The agent's goal is to explore the environment and gather information subject to its resource constraints in unknown, partially observable environments. Previous work has focused on the less general Adaptive Informative Path Planning (AIPP) problem, which considers only the effect of the agent's movement on received observations. The AIPPMS problem adds additional complexity by requiring that the agent reaso…
Navigating Demand Uncertainty in Container Shipping: Deep Reinforcement Learning for Enabling Adaptive and Feasible Master Stowage Planning
IT University of Copenhagen, Denmark（丹麦哥本哈根IT大学）；HEC Montréal, Qc, Canada（加拿大魁北克省蒙特利尔高等商学院）；GERAD, Qc, Canada（加拿大魁北克省GERAD）；Mila - Quebec AI Institute, Qc, Canada（加拿大魁北克省Mila-魁北克人工智能研究所）
Aerial robots are increasingly being utilized for environmental monitoring and exploration. However, a key challenge is efficiently planning paths to maximize the information value of acquired data as an initially unknown environment is explored. To address this, we propose a new approach for informative path planning based on deep reinforcement learning (RL). Combining recent advances in RL and robotic applications, our method combines tree search with an offline-learned neural network predicting informative sensing actions. We introduce several components making our approach applicable for r…
Why Reasoning Fails to Plan: A Planning-Centric Analysis of Long-Horizon Decision Making in LLM Agents
University of Notre Dame（诺丁汉大学）；Stanford University（斯坦福大学）；Purdue University（普渡大学）；University of Edinburgh（爱丁堡大学）；University of Oxford（牛津大学）；Yale University（耶鲁大学）
Fast and Precise: Adjusting Planning Horizon with Adaptive Subgoal Search
University of Warsaw（华沙大学）；IDEAS NCBR；Jagiellonian University（雅盖隆大学）；KAUST（阿卜杜拉国王科技大学）；Google Research（谷歌研究院）；Stanford University（斯坦福大学）；Polish Academy of Sciences（波兰科学院）
We study learning control in an online reset-free lifelong learning scenario, where mistakes can compound catastrophically into the future and the underlying dynamics of the environment may change. Traditional model-free policy learning methods have achieved successes in difficult tasks due to their broad flexibility, but struggle in this setting, as they can activate failure modes early in their lifetimes which are difficult to recover from and face performance degradation as dynamics change. On the other hand, model-based planning methods learn and adapt quickly, but require prohibitive leve…
Recent research in robot exploration and mapping has focused on sampling environmental hotspot fields. This exploration task is formalized by Low, Dolan, and Khosla (2008) in a sequential decision-theoretic planning under uncertainty framework called MASP. The time complexity of solving MASP approximately depends on the map resolution, which limits its use in large-scale, high-resolution exploration and mapping. To alleviate this computational difficulty, this paper presents an information-theoretic approach to MASP (iMASP) for efficient adaptive path planning; by reformulating the cost-minimi…
Scale-free adaptive planning for deterministic dynamics & discounted rewards
University of California, Berkeley, USA（加州大学伯克利分校）；Noah's Ark Lab, Huawei Technologies, London, UK（华为技术伦敦诺亚实验室）；Adobe Research, San Jose, USA（Adobe研究实验室）；SequeL team, INRIA Lille - Nord Europe, France（INRIA里尔-北欧洲SequeL团队）
Technical University of Denmark（丹麦技术大学）；Skåne University Hospital（斯科讷大学医院）；Danish Research Centre for Magnetic Resonance（丹麦磁共振研究中心）；Copenhagen University Hospital Hvidovre（哥本哈根大学医院维兹奥勒医院）；Copenhagen University Hospital Rigshospitalet（哥本哈根大学医院国立医院）；University College London（伦敦大学学院）；UCLH NHS Foundation Trust（伦敦大学学院医院NHS基金会信托）；National Hospital for Neurology and Neurosurgery（国立神经病学与神经外科医院）
In this paper we present a method for simultaneously segmenting brain tumors and an extensive set of organs-at-risk for radiation therapy planning of glioblastomas. The method combines a contrast-adaptive generative model for whole-brain segmentation with a new spatial regularization model of tumor shape using convolutional restricted Boltzmann machines. We demonstrate experimentally that the method is able to adapt to image acquisitions that differ substantially from any available training data, ensuring its applicability across treatment sites; that its tumor segmentation accuracy is compara…
