On the role of planning in model-based deep reinforcement learning
DeepMind（深度思维公司）
Jessica B. Hamrick, Abram L. Friesen, Feryal Behbahani, Arthur Guez, Fabio Viola, Sims Witherspoon, Thomas Anthony, Lars Buesing, Petar Veličković, Théophane Weber
Model-based planning is often thought to be necessary for deep, careful reasoning and generalization in artificial agents. While recent successes of model-based reinforcement learning (MBRL) with deep function approximation have strengthened this hypothesis, the resulting diversity of model-based methods has also made it difficult to track which components drive success and why. In this paper, we seek to disentangle the contributions of recent methods by focusing on three questions: (1) How does planning benefit MBRL agents? (2) Within planning, what choices drive performance? (3) To what extent does planning improve generalization? To answer these questions, we study the performance of MuZero (Schrittwieser et al., 2019), a state-of-the-art MBRL algorithm with strong connections and overlapping components with many other MBRL algorithms. We perform a number of interventions and ablations of MuZero across a wide range of environments, including control tasks, Atari, and 9x9 Go. Our results suggest the following: (1) Planning is most useful in the learning process, both for policy updates and for providing a more useful data distribution. (2) Using shallow trees with simple Monte-Carlo rollouts is as performant as more complex methods, except in the most difficult reasoning tasks. (3) Planning alone is insufficient to drive strong generalization. These results indicate where and how to utilize planning in reinforcement learning settings, and highlight a number of open questions for future MBRL research.
In agent control issues, the idea of combining reinforcement learning and planning has attracted much attention. Two methods focus on micro and macro action respectively. Their advantages would show together if there is a good cooperation between them. An essential for the cooperation is to find an appropriate boundary, assigning different functions to each method. Such boundary could be represented by parameters in a planning algorithm. In this paper, we create an optimization strategy for planning parameters, through analysis to the connection of reaction and planning; we also create a non-g…
TU Delft（代尔夫特理工大学）；Leiden University（莱顿大学）；Southeast University（东南大学）
Planning and reinforcement learning are two key approaches to sequential decision making. Multi-step approximate real-time dynamic programming, a recently successful algorithm class of which AlphaZero [Silver et al., 2018] is an example, combines both by nesting planning within a learning loop. However, the combination of planning and learning introduces a new question: how should we balance time spend on planning, learning and acting? The importance of this trade-off has not been explicitly studied before. We show that it is actually of key importance, with computational results indicating th…
Technion - Israel Institute of Technology（以色列理工学院）
A hallmark of intelligence is the ability to deduce general principles from examples, which are correct beyond the range of those observed. Generalized Planning deals with finding such principles for a class of planning problems, so that principles discovered using small instances of a domain can be used to solve much larger instances of the same domain. In this work we study the use of Deep Reinforcement Learning and Graph Neural Networks to learn such generalized policies and demonstrate that they can generalize to instances that are orders of magnitude larger than those they were trained on…
McGill University（麦吉尔大学）；Mila；Université de Montréal（蒙特利尔大学）；DeepMind
We present an end-to-end, model-based deep reinforcement learning agent which dynamically attends to relevant parts of its state during planning. The agent uses a bottleneck mechanism over a set-based representation to force the number of entities to which the agent attends at each planning step to be small. In experiments, we investigate the bottleneck mechanism with several sets of customized environments featuring different challenges. We consistently observe that the design allows the planning agents to generalize their learned task-solving abilities in compatible unseen environments by at…
INESC-ID / Instituto Superior Técnico（INESC-ID / 里斯本高等理工学院）
The model-based reinforcement learning paradigm, which uses planning algorithms and neural network models, has recently achieved unprecedented results in diverse applications, leading to what is now known as deep reinforcement learning. These agents are quite complex and involve multiple components, factors that can create challenges for research. In this work, we propose a new modular software architecture suited for these types of agents, and a set of building blocks that can be easily reused and assembled to construct new model-based reinforcement learning agents. These building blocks incl…
Despite significant progress in general AI planning, certain domains remain out of reach of current AI planning systems. Sokoban is a PSPACE-complete planning task and represents one of the hardest domains for current AI planners. Even domain-specific specialized search methods fail quickly due to the exponential search complexity on hard instances. Our approach based on deep reinforcement learning augmented with a curriculum-driven method is the first one to solve hard instances within one day of training while other modern solvers cannot solve these instances within any reasonable time limit…
Reinforcement learning and symbolic planning have both been used to build intelligent autonomous agents. Reinforcement learning relies on learning from interactions with real world, which often requires an unfeasibly large amount of experience. Symbolic planning relies on manually crafted symbolic knowledge, which may not be robust to domain uncertainties and changes. In this paper we present a unified framework {\em PEORL} that integrates symbolic planning with hierarchical reinforcement learning (HRL) to cope with decision-making in a dynamic environment with uncertainties. Symbolic plans ar…
In this paper a deep reinforcement based multi-agent path planning approach is introduced. The experiments are realized in a simulation environment and in this environment different multi-agent path planning problems are produced. The produced problems are actually similar to a vehicle routing problem and they are solved using multi-agent deep reinforcement learning. In the simulation environment, the model is trained on different consecutive problems in this way and, as the time passes, it is observed that the model's performance to solve a problem increases. Always the same simulation enviro…
By planning through a learned dynamics model, model-based reinforcement learning (MBRL) offers the prospect of good performance with little environment interaction. However, it is common in practice for the learned model to be inaccurate, impairing planning and leading to poor performance. This paper aims to improve planning with an importance sampling framework that accounts and corrects for discrepancy between the true and learned dynamics. This framework also motivates an alternative objective for fitting the dynamics model: to minimize the variance of value estimation during planning. We d…
Leiden Institute of Advanced Computer Science, Leiden University（莱顿大学莱顿高级计算机科学研究所）；TU Delft（代尔夫特理工大学）
Sequential decision making, commonly formalized as optimization of a Markov Decision Process, is a key challenge in artificial intelligence. Two successful approaches to MDP optimization are reinforcement learning and planning, which both largely have their own research communities. However, if both research fields solve the same problem, then we might be able to disentangle the common factors in their solution approaches. Therefore, this paper presents a unifying algorithmic framework for reinforcement learning and planning (FRAP), which identifies underlying dimensions on which MDP planning …
Reinforcement learning is an appropriate and successful method to robustly perform low-level robot control under noisy conditions. Symbolic action planning is useful to resolve causal dependencies and to break a causally complex problem down into a sequence of simpler high-level actions. A problem with the integration of both approaches is that action planning is based on discrete high-level action- and state spaces, whereas reinforcement learning is usually driven by a continuous reward function. However, recent advances in reinforcement learning, specifically, universal value function approx…
Although reinforcement learning has seen tremendous success recently, this kind of trial-and-error learning can be impractical or inefficient in complex environments. The use of demonstrations, on the other hand, enables agents to benefit from expert knowledge rather than having to discover the best action to take through exploration. In this survey, we discuss the advantages of using demonstrations in sequential decision making, various ways to apply demonstrations in learning-based decision making paradigms (for example, reinforcement learning and planning in the learned models), and how to …
Rutgers University, The State University of New Jersey（罗格斯大学（新泽西州立大学））
The aim of path planning is to reach the goal from starting point by searching for the route of an agent. In the path planning, the routes may vary depending on the number of variables such that it is important for the agent to reach various goals. Numerous studies, however, have dealt with a single goal that is predefined by the user. In the present study, I propose a novel reinforcement learning framework for a fully controllable agent in the path planning. To do this, I propose a bi-directional memory editing to obtain various bi-directional trajectories of the agent, in which the behavior …
Width-based algorithms search for solutions through a general definition of state novelty. These algorithms have been shown to result in state-of-the-art performance in classical planning, and have been successfully applied to model-based and model-free settings where the dynamics of the problem are given through simulation engines. Width-based algorithms performance is understood theoretically through the notion of planning width, providing polynomial guarantees on their runtime and memory consumption. To facilitate synergies across research communities, this paper summarizes the area of widt…
Courant Institute of Mathematical Sciences, New York University（纽约大学库朗数学科学研究所）；New York University（纽约大学）
Reinforcement learning algorithms have had tremendous successes in online learning settings. However, these successes have relied on low-stakes interactions between the algorithmic agent and its environment. In many settings where RL could be of use, such as health care and autonomous driving, the mistakes made by most online RL algorithms during early training come with unacceptable costs. These settings require developing reinforcement learning algorithms that can operate in the so-called batch setting, where the algorithms must learn from set of data that is fixed, finite, and generated fro…
We introduce a learning method called ``gradient-based reinforcement planning'' (GREP). Unlike traditional DP methods that improve their policy backwards in time, GREP is a gradient-based method that plans ahead and improves its policy before it actually acts in the environment. We derive formulas for the exact policy gradient that maximizes the expected future reward and confirm our ideas with numerical experiments.
National University of Ireland Galway（爱尔兰国立高威大学）；Vrije Universiteit Brussel（布鲁塞尔自由大学）；Linköping University（林雪平大学）；University of Amsterdam（阿姆斯特丹大学）；University of Oxford（牛津大学）；Deakin University（迪肯大学）；University of Washington（华盛顿大学）；Universidade do Vale do Rio dos Sinos（南里奥格兰德河谷大学）；Politecnico di Milano（米兰理工大学）；Federation University Australia（澳大利亚联邦大学）
Real-world decision-making tasks are generally complex, requiring trade-offs between multiple, often conflicting, objectives. Despite this, the majority of research in reinforcement learning and decision-theoretic planning either assumes only a single objective, or that multiple objectives can be adequately handled via a simple linear combination. Such approaches may oversimplify the underlying problem and hence produce suboptimal results. This paper serves as a guide to the application of multi-objective methods to difficult problems, and is aimed at researchers who are already familiar with …
In this paper, we tackle the problem of Unmanned Aerial (UA V) path planning in complex and uncertain environments by designing a Model Predictive Control (MPC), based on a Long-Short-Term Memory (LSTM) network integrated into the Deep Deterministic Policy Gradient algorithm. In the proposed solution, LSTM-MPC operates as a deterministic policy within the DDPG network, and it leverages a predicting pool to store predicted future states and actions for improved robustness and efficiency. The use of the predicting pool also enables the initialization of the critic network, leading to improved co…
MIT（麻省理工学院）；MIT-IBM Watson AI Lab（麻省理工学院-IBM沃森人工智能实验室）；IBM Research（IBM研究院）
Recent advances in reinforcement learning (RL) have led to a growing interest in applying RL to classical planning domains or applying classical planning methods to some complex RL domains. However, the long-horizon goal-based problems found in classical planning lead to sparse rewards for RL, making direct application inefficient. In this paper, we propose to leverage domain-independent heuristic functions commonly used in the classical planning literature to improve the sample efficiency of RL. These classical heuristics act as dense reward generators to alleviate the sparse-rewards issue an…
In Meta-Reinforcement Learning (meta-RL) an agent is trained on a set of tasks to prepare for and learn faster in new, unseen, but related tasks. The training tasks are usually hand-crafted to be representative of the expected distribution of test tasks and hence all used in training. We show that given a set of training tasks, learning can be both faster and more effective (leading to better performance in the test tasks), if the training tasks are appropriately selected. We propose a task selection algorithm, Information-Theoretic Task Selection (ITTS), based on information theory, which opt…
NEC Corporation（日本电气股份有限公司）；National Institute of Advanced Industrial Science and Technology（产业技术综合研究所）；The University of Tokyo（东京大学）
One of the key challenges in applying reinforcement learning to real-life problems is that the amount of train-and-error required to learn a good policy increases drastically as the task becomes complex. One potential solution to this problem is to combine reinforcement learning with automated symbol planning and utilize prior knowledge on the domain. However, existing methods have limitations in their applicability and expressiveness. In this paper we propose a hierarchical reinforcement learning method based on abductive symbolic planning. The planner can deal with user-defined evaluation fu…
University of Alberta（阿尔伯塔大学）；ETH Zurich（苏黎世联邦理工学院）；University College London（伦敦大学学院）；DeepMind（深度思维）
A practical challenge in reinforcement learning are combinatorial action spaces that make planning computationally demanding. For example, in cooperative multi-agent reinforcement learning, a potentially large number of agents jointly optimize a global reward function, which leads to a combinatorial blow-up in the action space by the number of agents. As a minimal requirement, we assume access to an argmax oracle that allows to efficiently compute the greedy policy for any Q-function in the model class. Building on recent work in planning with local access to a simulator and linear function ap…
With the impact of real-time processing being realized in the recent past, the need for efficient implementations of reinforcement learning algorithms has been on the rise. Albeit the numerous advantages of Bellman equations utilized in RL algorithms, they are not without the large search space of design parameters. This research aims to shed light on the design space exploration associated with reinforcement learning parameters, specifically that of Policy Iteration. Given the large computational expenses of fine-tuning the parameters of reinforcement learning algorithms, we propose an auto-t…
Sequential decision making, commonly formalized as Markov Decision Process (MDP) optimization, is a important challenge in artificial intelligence. Two key approaches to this problem are reinforcement learning (RL) and planning. This paper presents a survey of the integration of both fields, better known as model-based reinforcement learning. Model-based RL has two main steps. First, we systematically cover approaches to dynamics model learning, including challenges like dealing with stochasticity, uncertainty, partial observability, and temporal abstraction. Second, we present a systematic ca…
Humans learn to play video games significantly faster than the state-of-the-art reinforcement learning (RL) algorithms. People seem to build simple models that are easy to learn to support planning and strategic exploration. Inspired by this, we investigate two issues in leveraging model-based RL for sample efficiency. First we investigate how to perform strategic exploration when exact planning is not feasible and empirically show that optimistic Monte Carlo Tree Search outperforms posterior sampling methods. Second we show how to learn simple deterministic models to support fast learning usi…
Offline reinforcement learning (RL) provides a framework for learning decision-making from offline data and therefore constitutes a promising approach for real-world applications as automated driving. Self-driving vehicles (SDV) learn a policy, which potentially even outperforms the behavior in the sub-optimal data set. Especially in safety-critical applications as automated driving, explainability and transferability are key to success. This motivates the use of model-based offline RL approaches, which leverage planning. However, current state-of-the-art methods often neglect the influence of…
The history of learning for control has been an exciting back and forth between two broad classes of algorithms: planning and reinforcement learning. Planning algorithms effectively reason over long horizons, but assume access to a local policy and distance metric over collision-free paths. Reinforcement learning excels at learning policies and the relative values of states, but fails to plan over long horizons. Despite the successes of each method in various domains, tasks that require reasoning over long horizons with limited feedback and high-dimensional observations remain exceedingly chal…
Okinawa Institute of Science and Technology（冲绳科学技术大学院大学）
What is the difference between goal-directed and habitual behavior? We propose a novel computational framework of decision making with Bayesian inference, in which everything is integrated as an entire neural network model. The model learns to predict environmental state transitions by self-exploration and generating motor actions by sampling stochastic internal states ${z}$. Habitual behavior, which is obtained from the prior distribution of ${z}$, is acquired by reinforcement learning. Goal-directed behavior is determined from the posterior distribution of ${z}$ by planning, using active inf…
Identifying uncertainty and taking mitigating actions is crucial for safe and trustworthy reinforcement learning agents, especially when deployed in high-risk environments. In this paper, risk sensitivity is promoted in a model-based reinforcement learning algorithm by exploiting the ability of a bootstrap ensemble of dynamics models to estimate environment epistemic uncertainty. We propose uncertainty guided cross-entropy method planning, which penalises action sequences that result in high variance state predictions during model rollouts, guiding the agent to known areas of the state space w…
Coverage path planning in a generic known environment is shown to be NP-hard. When the environment is unknown, it becomes more challenging as the robot is required to rely on its online map information built during coverage for planning its path. A significant research effort focuses on designing heuristic or approximate algorithms that achieve reasonable performance. Such algorithms have sub-optimal performance in terms of covering the area or the cost of coverage, e.g., coverage time or energy consumption. In this paper, we provide a systematic analysis of the coverage problem and formulate …
State-of-the-art efficient model-based Reinforcement Learning (RL) algorithms typically act by iteratively solving empirical models, i.e., by performing \emph{full-planning} on Markov Decision Processes (MDPs) built by the gathered experience. In this paper, we focus on model-based RL in the finite-state finite-horizon MDP setting and establish that exploring with \emph{greedy policies} -- act by \emph{1-step planning} -- can achieve tight minimax performance in terms of regret, $\tilde{\mathcal{O}}(\sqrt{HSAT})$. Thus, full-planning in model-based RL can be avoided altogether without any perf…
Volvo Group（沃尔沃集团）；Chalmers University of Technology（查尔姆斯理工大学）；University of Illinois at Urbana-Champaign（伊利诺伊大学厄巴纳-香槟分校）；Stanford University（斯坦福大学）
Tactical decision making for autonomous driving is challenging due to the diversity of environments, the uncertainty in the sensor information, and the complex interaction with other road users. This paper introduces a general framework for tactical decision making, which combines the concepts of planning and learning, in the form of Monte Carlo tree search and deep reinforcement learning. The method is based on the AlphaGo Zero algorithm, which is extended to a domain with a continuous state space where self-play cannot be used. The framework is applied to two different highway driving cases …
In the world of advice and financial planning, there is seldom one right answer. While traditional algorithms have been successful in solving linear problems, its success often depends on choosing the right features from a dataset, which can be a challenge for nuanced financial planning scenarios. Reinforcement learning is a machine learning approach that can be employed with complex data sets where picking the right features can be nearly impossible. In this paper, we will explore the use of machine learning for financial forecasting, predicting economic indicators, and creating a savings str…
Fraunhofer Institute for Integrated Circuits IIS（弗劳恩霍夫集成电路研究所 IIS）；Friedrich-Alexander-University Erlangen-Nuremberg (FAU)（弗里德里希-亚历山大-埃尔兰根-纽伦堡大学（FAU））
This paper presents a novel motion and trajectory planning algorithm for nonholonomic mobile robots that uses recent advances in deep reinforcement learning. Starting from a random initial state, i.e., position, velocity and orientation, the robot reaches an arbitrary target state while taking both kinematic and dynamic constraints into account. Our deep reinforcement learning agent not only processes a continuous state space it also executes continuous actions, i.e., the acceleration of wheels and the adaptation of the steering angle. We evaluate our motion and trajectory planning on a mobile…
In this paper, we study the problem of deceptive reinforcement learning to preserve the privacy of a reward function. Reinforcement learning is the problem of finding a behaviour policy based on rewards received from exploratory behaviour. A key ingredient in reinforcement learning is a reward function, which determines how much reward (negative or positive) is given and when. However, in some situations, we may want to keep a reward function private; that is, to make it difficult for an observer to determine the reward function used. We define the problem of privacy-preserving reinforcement l…
Deep reinforcement learning (DRL) has gained great success by learning directly from high-dimensional sensory inputs, yet is notorious for the lack of interpretability. Interpretability of the subtasks is critical in hierarchical decision-making as it increases the transparency of black-box-style DRL approach and helps the RL practitioners to understand the high-level behavior of the system better. In this paper, we introduce symbolic planning into DRL and propose a framework of Symbolic Deep Reinforcement Learning (SDRL) that can handle both high-dimensional sensory inputs and symbolic planni…
Provable Representation with Efficient Planning for Partial Observable Reinforcement Learning
University of Alberta（阿尔伯塔大学）；UT Austin（德克萨斯大学奥斯汀分校）；The Chinese University of Hong Kong, Shenzhen（香港中文大学（深圳））；Google DeepMind（谷歌DeepMind）；Georgia Tech（佐治亚理工学院）
