Reinforcement learning and human behavior

To behave adaptively, we must learn from the consequences of our actions. Doing so is difficult when the consequences of an action follow a delay. This introduces the problem of temporal credit assignment. When feedback follows a sequence of decisions, how should the individual assign credit to the intermediate actions that comprise the sequence? Research in reinforcement learning provides 2 general solutions to this problem: model-free reinforcement learning and model-based reinforcement learning. In this review, we examine connections between stimulus-response and cognitive learning theories, habitual and goal-directed control, and model-free and model-based reinforcement learning. We then consider a range of problems related to temporal credit assignment. These include second-order conditioning and secondary reinforcers, latent learning and detour behavior, partially observable Markov decision processes, actions with distributed outcomes, and hierarchical learning. We ask whether humans and animals, when faced with these problems, behave in a manner consistent with reinforcement learning techniques. Throughout, we seek to identify neural substrates of model-free and model-based reinforcement learning. The former class of techniques is understood in terms of the neurotransmitter dopamine and its effects in the basal ganglia. The latter is understood in terms of a distributed network of regions including the prefrontal cortex, medial temporal lobes, cerebellum, and basal ganglia. Not only do reinforcement learning techniques have a natural interpretation in terms of human and animal behavior but they also provide a useful framework for understanding neural reward valuation and action selection.

Recent evidence in neuroscience and psychology suggests that a single reinforcement learning (RL) algorithm only accounts for less than 60% of the variance of human choice behavior in an uncertain and dynamic environment, where the amount of uncertainty in state-action-state transitions drift over time. The prediction performance further decreases when the size of the state space increases. We proposed a hierarchical context-dependent RL control framework that dynamically exerted control weights on model-based (MB) and multiple model-free (MF) RL strategies associated with different task goals. To properly assess the validity of the proposed method, we considered a two-stage Markov decision task (MDT) in which the three different types of context changed over time. We trained 57 different RL control models on a Caltech MDT data set; then, we assessed their prediction performance using a Bayesian model comparison. This large-scale computer simulation analysis revealed that the model providing the most accurate prediction was the version that implemented the competition between the MB and multiple goal-dependent MF RL strategies. The present study demonstrates the applicability of the goal-driven RL control to a variety of real-world human-robot interaction scenarios.

Reinforcement learning (RL) is one of the popular methods for intelligent control and decision making in the field of robotics recently. The goal of RL is to learn an optimal policy of the agent by interacting with the environment via trail and error. There are two main algorithms for RL problems, including model-free and model-based methods. Model-free RL is driven by historical trajectories and empirical data of the agent to optimize the policy, which needs to take actions in the environment to collect the trajectory data and may cause the damage of the robot during training in the real environment. The main different between model-based and model-free RL is that a model of the transition probability in the interaction environment is employed. Thus the agent can search the optimal policy through internal simulation. However, the model of the transition probability is usually estimated from historical data in a single environment with statistical errors. Therefore, an issue is faced by the agent is that the optimal policy is sensitive to perturbations in the model of the environment which can lead to serious degradation in performance. Robust RL aims to learn a robust optimal policy that accounts for model uncertainty of the transition probability to systematically mitigate the sensitivity of the optimal policy in perturbed environments. In this overview, we begin with an introduction to the algorithms in RL, then focus on the model uncertainty of the transition probability in robust RL. In parallel, we highlight the current research and challenges of robust RL for robot control. To conclude, we describe some research areas in robust RL and look ahead to the future work about robot control in complex environments.

Despite the fact that deep reinforcement learning (RL) has surpassed human-level performances in various tasks, it still has several fundamental challenges. First, most RL methods require intensive data from the exploration of the environment to achieve satisfactory performance. Second, the use of neural networks in RL renders it hard to interpret the internals of the system in a way that humans can understand. To address these two challenges, we propose a framework that enables an RL agent to reason over its exploration process and distill high-level knowledge for effectively guiding its future explorations. Specifically, we propose a novel RL algorithm that learns high-level knowledge in the form of a finite reward automaton by using the L* learning algorithm. We prove that in episodic RL, a finite reward automaton can express any non-Markovian bounded reward functions with finitely many reward values and approximate any non-Markovian bounded reward function (with infinitely many reward values) with arbitrary precision. We also provide a lower bound for the episode length such that the proposed RL approach almost surely converges to an optimal policy in the limit. We test this approach on two RL environments with non-Markovian reward functions, choosing a variety of tasks with increasing complexity for each environment. We compare our algorithm with the state-of-the-art RL algorithms for non-Markovian reward functions, such as Joint Inference of Reward machines and Policies for RL (JIRP), Learning Reward Machine (LRM), and Proximal Policy Optimization (PPO2). Our results show that our algorithm converges to an optimal policy faster than other baseline methods.

Animals integrate knowledge about how the state of the environment evolves to choose actions that maximise reward. Such goal-directed behaviour - or model-based (MB) reinforcement learning (RL) - can flexibly adapt choice to changes, being thus distinct from simpler habitual - or model-free (MF) RL - strategies. Previous inactivation and neuroimaging work implicates prefrontal cortex (PFC) and the caudate striatal region in MB-RL; however, details are scarce about its implementation at the single-neuron level. Here, we recorded from two PFC regions – the dorsal anterior cingulate cortex (ACC) and dorsolateral PFC (DLPFC), and two striatal regions, caudate and putamen – while two rhesus macaques performed a sequential decision-making (two-step) task in which MB-RL involves knowledge about the statistics of reward and state transitions. All four regions, but particularly the ACC, encoded the rewards received and tracked the probabilistic state transitions that occurred. However, ACC (and to a lesser extent caudate) encoded the key variables of the task - namely the interaction between reward, transition and choice – which underlies MB decision-making. ACC and caudate neurons also encoded MB-derived estimates of choice values. Moreover, caudate value estimates of the choice options flipped when a rare transition occurred, demonstrating value update based on structural knowledge of the task. The striatal regions were unique (relative to PFC) in encoding the current and previous rewards with opposing polarities, reminiscent of dopaminergic neurons, and indicative of a MF prediction error. Our findings provide a deeper understanding of selective and temporally dissociable neural mechanisms underlying goal-directed behaviour.

Model-free episodic reinforcement learning problems define the environment reward with functions that often provide only sparse information throughout the task. Consequently, agents are not given enough feedback about the fitness of their actions until the task ends with success or failure. Previous work addresses this problem with reward shaping. In this paper we introduce a novel approach to improve model-free reinforcement learning agents' performance with a three step approach. Specifically, we collect demonstration data, use the data to recover a linear function using inverse reinforcement learning and we use the recovered function for potential-based reward shaping. Our approach is model-free and scalable to high dimensional domains. To show the scalability of our approach we present two sets of experiments in a two dimensional Maze domain, and the 27 dimensional Mario AI domain. We compare the performance of our algorithm to previously introduced reinforcement learning from demonstration algorithms. Our experiments show that our approach outperforms the state-of-the-art in cumulative reward, learning rate and asymptotic performance.

Reinforcement learning (RL) focuses on an essential aspect of intelligent behavior – how an agent can learn to make good decisions given experience and rewards in a stochastic<br>world. Yet popular RL algorithms that have enabled exciting successes in domains with good simulators (Go, Atari, etc) still often fail to learn in other domains because they rely on<br>simple heuristics for exploration. This provides additional empirical justification for essential questions around RL, specifically around algorithms that learn in a provably efficient manner through strategic exploration in any considered domain. This thesis provides new algorithms<br>and theory that enable good performance with respect to existing theoretical frameworks for evaluating RL algorithms (specifically, probably approximately correct) and introduces new stronger evaluation criteria, that may be particularly of interest as RL is applied to more real world problems.<br>For the first line of work on probably approximately correct (PAC) RL algorithms, we introduce a series of algorithms for episodic tabular domains with substantially better PAC<br>sample complexity bounds that culminate in a new algorithm with close to minimax optimal PAC and regret bounds. Look up tables are required by most sample efficient and computationally tractable algorithms, but cannot represent many practical domains. We therefore also present a new RL algorithm that can learn a good policy in environments with high dimensional observations and hidden deterministic states; unlike predecessors, this algorithm provably<br>explores not only in a statistically but also computationally efficient manner assuming access to function classes with efficient optimization oracles. To make progress it is critical to have the right measures of success. While empirical<br>demonstrations are quite clear, we find that for theoretical properties, two of the most commonly used learning frameworks, PAC guarantees and regret guarantees, each allow undesirable algorithm behavior (e.g. ignoring new observations that could improve the policy). We present<br>a new stronger learning framework called Uniform-PAC that unifies the existing frameworks and prevents undesirable algorithm properties. One caveat of all existing learning frameworks is that for any particular episode, we do not<br>know how well the algorithm will perform. To address this, we introduce the IPOC framework that requires algorithms to provide a certificate before each episode bounding how suboptimal the current policy can be. Such certifications may be of substantial interest in high stakes scenarios when an organization may wish to track or even pause an online RL system should the potential expected performance bound drop below a required expected outcome.

Deep reinforcement learning (RL), where the agent learns from mistakes, has been successfully applied to a variety of tasks. With the aim of learning collision-free policies for unmanned vehicles, deep RL has been used for training with various types of data, such as colored images, depth images, and LiDAR point clouds, without the use of classic map--localize--plan approaches. However, existing methods are limited by their reliance on cameras and LiDAR devices, which have degraded sensing under adverse environmental conditions (e.g., smoky environments). In response, we propose the use of single-chip millimeter-wave (mmWave) radar, which is lightweight and inexpensive, for learning-based autonomous navigation. However, because mmWave radar signals are often noisy and sparse, we propose a cross-modal contrastive learning for representation (CM-CLR) method that maximizes the agreement between mmWave radar data and LiDAR data in the training stage. We evaluated our method in real-world robot compared with 1) a method with two separate networks using cross-modal generative reconstruction and an RL policy and 2) a baseline RL policy without cross-modal representation. Our proposed end-to-end deep RL policy with contrastive learning successfully navigated the robot through smoke-filled maze environments and achieved better performance compared with generative reconstruction methods, in which noisy artifact walls or obstacles were produced. All pretrained models and hardware settings are open access for reproducing this study and can be obtained at https://arg-nctu.github.io/projects/deeprl-mmWave.html

This paper introduces a visual tracking control scheme for the two axes Inertially Stabilized Platform. An adaptive PID controller based on the Deep Deterministic Policy Gradient (DDPG) algorithm is used where it is treated as a Reinforcement Learning Policy. The optimal PID parameters are obtained at each time step via a training process. To effectively train the parameters, we address the problem of handling time delay and uncertainty of the system by presenting an observer system. Simulation results show that our method achieves good performance for various cases of time delay.

Arbitration between model-based (MB) and model-free (MF) reinforcement learning (RL) is key feature of human reinforcement learning. The computational model of arbitration control has been demonstrated to outperform conventional reinforcement learning algorithm, in terms of not only behavioral data but also neural signals. However, this arbitration process does not take full account of contextual changes in environment during learning. By incorporating a Dirichlet process Gaussian mixture model into the arbitration process, we propose a meta-controller for RL that quickly adapts to contextual changes of environment. The proposed model performs better than a conventional model-free RL, model-based RL, and arbitration model.

Introduction Reinforcement learning formalizes the concept of learning from interactions.1 Broadly, reinforcement learning focuses on a setting in which an agent (decision maker) sequentially interacts with an environment that is partially unknown to them. At each stage, the agent takes an action and receives a reward. The objective of the agent is to maximize rewards accumulated in the long run. There are many situations in health care where decisions are made sequentially for which reinforcement learning approaches could prove useful for decision making. Throughout this article, we consider treatment prescription as an archetypical example to connect reinforcement learning concepts to a health care setting. In this setting, the care provider, the prescribed treatment, and the patients can be viewed as the agent, the action, and the environment, respectively, as depicted in Figure 1.Figure 1: Sequential treatment of AKI or CKD complications modeled as a reinforcement learning problem.Background In this section, with the objective of making reinforcement learning literature more accessible to a clinical audience, we briefly introduce related fundamental concepts and approaches. We refer the interested reader to Sutton and Barto1 for a comprehensive introduction to reinforcement learning. Markov Decision Processes Markov decision processes (MDPs) are a formalism of the sequential decision-making problem that has been central to the theoretical and practical advancements of reinforcement learning. In each stage of an MDP, the agent observes the state of the environment and takes an action, which, in turn, results in a change of the state. This change of state is assumed to be probabilistic with the next state being determined only by the preceding state, the chosen action, and the transition probability. The agent also receives a reward that is a function of the taken action, the preceding state, and the subsequent state. In an MDP, the objective of the agent is to maximize the return defined as the reward accumulated over a time horizon. In some applications, it is common to consider the horizon to be infinite, in which case the future rewards are discounted by a factor smaller than one. The selection of action by the agent on the basis of the observed state is known as the policy. More formally, a policy is a probabilistic mapping from states to each possible action. Because the policy and the reward are a function of the state, it is critical to estimate the utility of being in a certain state. More specifically, the value function is defined as the expected return starting from a given state under the chosen policy. Under this formalism, the objective of the agent is to find the optimal policy that maximizes the value function for all states. Reinforcement Learning Methods Action-value methods are a class of reinforcement learning methods in which the actions are chosen on the basis of the estimation of their long-term value. A prominent example of an action-value method is Q-learning in which the agent iteratively takes actions with the highest estimated values and updates the action-state value function on the basis of new observations. Policy gradient methods are another class of reinforcement learning methods that seek to optimize the policy directly instead of choosing actions on the basis of their respective estimated value. Such methods could be advantageous in health care applications that entail a large number of possible actions, e.g., when recommending a wide range of drug dosages or treatment options. Clinical Applications Reinforcement learning frameworks and methods are broadly applicable to clinical settings in which decisions are made sequentially. A prominent clinical application of reinforcement learning is for treatment recommendation, which has been studied across a variety of diseases and treatments including radiation and chemotherapy for cancer, brain stimulation for epilepsy, and treatment strategies for sepsis.2–5 In such treatment recommendation settings, a policy is commonly known as a dynamic treatment regime. There are various other clinical applications of reinforcement learning including diagnosis, medical imaging, and decision support tools (see refs. 2–5 and the references therein). Reinforcement Learning in Nephrology Although there have been recent applications of machine learning in nephrology,6,7 to the best of the authors' knowledge, the application of reinforcement learning to nephrology has been primarily limited to optimizing the erythropoietin dosage in hemodialysis patients.8,9 However, there are other settings where reinforcement learning has the potential to improve patient care in nephrology. For example, reinforcement learning methods can be adopted in the treatment of the complications of AKI or CKD (Figure 1). In this problem, the state models the conditions of the patient (e.g., vital signs, laboratory test results including urine and blood tests, and urine output measurements). The action refers to the treatment options (e.g., the dosage of medications such as sodium polystyrene sulfonate, and hemodialysis). The reward models the improvement in patient conditions. Similarly, reinforcement learning can help automate and optimize the dosage of immunosuppressive drugs in kidney transplants. Challenges and Opportunities Despite the success of reinforcement learning in several simplified clinical settings, their large-scale application to patient care faces several open challenges. The complexity of human biology complicates modeling clinical decision making as a reinforcement learning problem. The state space in such settings is often enormous, which could make a purely computational approach infeasible. Moreover, modeling all potential objectives a priori as a reward function may not be feasible. To overcome these challenges and realize the potential of reinforcement learning, clinical insight can play a pivotal role. More specifically, restricting the state space to only include highly relevant clinical variables could greatly reduce the computational complexity. Furthermore, using inverse reinforcement learning,2 relevant reward functions can be learned from retrospective studies assuming the optimality of clinical decisions. Another critical challenge is addressing moral and ethical concerns. It is imperative to ensure that reinforcement learning methods do not cause harm to the patient. To this end, there exists a need for a thorough validation of such methods before their use in patient care. Hence, there is a need to go beyond retrospective studies that have been used for the proof of concept of most existing reinforcement learning methods in health care applications.2,3 The lessons learned from the success of reinforcement learning in other application areas (e.g., self-driving cars) can help navigate the path to realizing its potential in health care. Accessible open-source simulation environments that enable researchers to compare various approaches are essential to the field of reinforcement learning. OpenAI Gym is currently the leading toolkit containing a wide range of simulated environments, e.g., surgical robotics.10 The development of high-quality and reliable simulation environments for nephrology and other health care applications can facilitate the development and validation of reinforcement learning methods beyond limited retrospective studies. The adoption of methods validated in such simulation environments in actual clinical settings will require clinicians' oversight. Similar to how self-driving cars require a human driver to ensure collision avoidance, clinicians' oversight is critical to ensure the safety of the patients, especially in the early stages of the adoption of reinforcement learning methods. The data from clinicians' decisions (e.g., overruling the automated treatment recommendation) can be used to improve the reliability of autonomous systems over time and reduce the burden of clinicians' oversight.

Deep reinforcement learning (RL) methods often require many trials before convergence, and no direct interpretability of trained policies is provided. In order to achieve fast convergence and interpretability for the policy in RL, we propose a novel RL method for text-based games with a recent neuro-symbolic framework called Logical Neural Network, which can learn symbolic and interpretable rules in their differentiable network. The method is first to extract first-order logical facts from text observation and external word meaning network (ConceptNet), then train a policy in the network with directly interpretable logical operators. Our experimental results show RL training with the proposed method converges significantly faster than other state-of-the-art neuro-symbolic methods in a TextWorld benchmark.

Fuzzy-based predictive deep reinforcement learning for robust and constrained optimal control of industrial solar thermal plants

Autonomous control in high-dimensional, continuous state spaces is a persistent and important challenge in the fields of robotics and artificial intelligence. Because of high risk and complexity, the adoption of AI for autonomous combat systems has been a long-standing difficulty. In order to address these issues, DARPA's AlphaDogfight Trials (ADT) program sought to vet the feasibility of and increase trust in AI for autonomously piloting an F-16 in simulated air-to-air combat. Our submission to ADT solves the high-dimensional, continuous control problem using a novel hierarchical deep reinforcement learning approach consisting of a high-level policy selector and a set of separately trained low-level policies specialized for excelling in specific regions of the state space. Both levels of the hierarchy are trained using off-policy, maximum entropy methods with expert knowledge integrated through reward shaping. Our approach outperformed human expert pilots and achieved a second-place rank in the ADT championship event. <italic xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">Impact Statement–</i> Significant performance milestones in reinforcement learning have been achieved in recent years, with autonomous agents demonstrating super-human performance across a wide variety of tasks. Before these algorithms can be extensively deployed in real-world defense applications, a greater level of trust must first be achieved. ADT was an important step towards developing the trust necessary to operationalize these algorithms, by demonstrating their effectiveness on a foundational yet relevant problem in a high-fidelity simulation environment. Developed for the program, our hierarchical reinforcement learning agent was designed alongside of and competed against active fighter pilots, and ultimately defeated a graduate of the United States Air Force's F-16 Weapons Instructor Course in match play.

Bidirectional Human-Robot Learning: Imitation and Skill Improvement