Muscle-actuated organisms are capable of learning an unparalleled diversity of dexterous movements despite their vast amount of muscles. Reinforcement learning (RL) on large musculoskeletal models, however, has not been able to show similar performance. We conjecture that ineffective exploration in large overactuated action spaces is a key problem. This is supported by the finding that common exploration noise strategies are inadequate in synthetic examples of overactuated systems. We identify differential extrinsic plasticity (DEP), a method from the domain of self-organization, as being able to induce state-space covering exploration within seconds of interaction. By integrating DEP into RL, we achieve fast learning of reaching and locomotion in musculoskeletal systems, outperforming current approaches in all considered tasks in sample efficiency and robustness.

The policy represented by the deep neural network can overfit the spurious features in observations, which hamper a reinforcement learning agent from learning effective policy. This issue becomes severe in high-dimensional state, where the agent struggles to learn a useful policy. Data augmentation can provide a performance boost to RL agents by mitigating the effect of overfitting. However, such data augmentation is a form of prior knowledge, and naively applying them in environments might worsen an agent's performance. In this paper, we propose a novel RL algorithm to mitigate the above issue and improve the efficiency of the learned policy. Our approach consists of a max-min game theoretic objective where a perturber network modifies the state to maximize the agent's probability of taking a different action while minimizing the distortion in the state. In contrast, the policy network updates its parameters to minimize the effect of perturbation while maximizing the expected future reward. Based on this objective, we propose a practical deep reinforcement learning algorithm, Adversarial Policy Optimization (APO). Our method is agnostic to the type of policy optimization, and thus data augmentation can be incorporated to harness the benefit. We evaluated our approaches on several DeepMind Control robotic environments with high-dimensional and noisy state settings. Empirical results demonstrate that our method APO consistently outperforms the state-of-the-art on-policy PPO agent. We further compare our method with state-of-the-art data augmentation, RAD, and regularization-based approach DRAC. Our agent APO shows better performance compared to these baselines.

Training DRL agents is often a time-consuming process as a large number of samples and environment interactions is required. This effect is even amplified in the case of Batch RL, where the agent is trained without environment interactions solely based on a set of previously collected data. Novel approaches based on quantum computing suggest an advantage compared to classical approaches in terms of sample efficiency. To investigate this advantage, we propose a batch RL algorithm leveraging VQC as function approximators in the discrete BCQ algorithm. Additionally, we present a novel data re-uploading scheme based on cyclically shifting the input variables' order in the data encoding layers. We show the efficiency of our algorithm on the OpenAI CartPole environment and compare its performance to classical neural network-based discrete BCQ.

ABSTRACT A substantial part of fighter pilot training is simulation-based and involves computer-generated forces controlled by predefined behavior models. The behavior models are typically manually created by eliciting knowledge from experienced pilots, which is a time-consuming process. Despite the work put in, the behavior models are often unsatisfactory due to their predictable nature and lack of adaptivity, forcing instructors to spend time manually monitoring and controlling them. Reinforcement and imitation learning pose as alternatives to handcrafted models. This paper presents the Learning Environment for the Air Domain (LEAD), a system for creating and integrating intelligent air combat behavior in military simulations. By incorporating the popular programming library and interface Gymnasium, LEAD allows users to apply readily available machine learning algorithms. Additionally, LEAD can communicate with third-party simulation software through distributed simulation protocols, which allows behavior models to be learned and employed using simulation systems of different fidelities. 1 INTRODUCTION A large part of the training fighter pilots undergo occurs in simulators under instructor supervision. In these simulators, the pilots practice tactics and operations by engaging in scenarios including friendly and hostile forces, often represented by computer-generated forces (CGFs), which are autonomous or semi-autonomous actors used in military simulation (Løvlid et al. 2017). These CGFs must behave in a way that accelerates training and builds the necessary competence of the pilots. Still, a current limitation to using CGFs for training is that their behaviors often come across as predictable, inviting pilots to exploit their vulnerabilities rather than focus on achieving the training objectives (Toubman 2020, ch. 1). Such constraints in the behavior models force instructors to micromanage the CGFs, restricting the complexity of scenarios that can be managed and trained (Källström et al. 2022). Besides, qualified instructors are both in short supply and on tight schedules, meaning they should devote their full attention to giving instructions and feedback to pilots. Modeling adaptive and intelligent air combat behavior for CGFs is thus a key challenge.

Artificial Intelligence and digital health have the potential to transform global health. However, having access to representative data to test and validate algorithms in realistic production environments is essential. We introduce HealthSyn, an open-source synthetic data generator of user behavior for testing reinforcement learning algorithms in the context of mobile health interventions. The generator utilizes Markov processes to generate diverse user actions, with individual user behavioral patterns that can change in reaction to personalized interventions (i.e., reminders, recommendations, and incentives). These actions are translated into actual logs using an ML-purposed data schema specific to the mobile health application functionality included with HealthKit, and open-source SDK. The logs can be fed to pipelines to obtain user metrics. The generated data, which is based on real-world behaviors and simulation techniques, can be used to develop, test, and evaluate, both ML algorithms in research and end-to-end operational RL-based intervention delivery frameworks.

This paper introduces a learning-based framework for robot adaptive manipulating the object with a revolute joint in unstructured environments. We concentrate our discussion on various cabinet door opening tasks. To improve the performance of Deep Reinforcement Learning in this scene, we analytically provide an efficient sampling manner utilizing the constraints of the objects. To open various kinds of doors, we add encoded environment parameters that define the various environments to the input of out policy. To transfer the policy into the real world, we train an adaptation module in simulation and fine-tune the adaptation module to cut down the impact of the policy-unaware environment parameters. We design a series of experiments to validate the efficacy of our framework. Additionally, we testify to the model's performance in the real world compared to the traditional door opening method.

Reinforcement learning (RL) allows an agent interacting sequentially with an environment to maximize its long-term expected return. In the distributional RL (DistrRL) paradigm, the agent goes beyond the limit of the expected value, to capture the underlying probability distribution of the return across all time steps. The set of DistrRL algorithms has led to improved empirical performance. Nevertheless, the theory of DistrRL is still not fully understood, especially in the control case. In this paper, we present the simpler one-step distributional reinforcement learning (OS-DistrRL) framework encompassing only the randomness induced by the one-step dynamics of the environment. Contrary to DistrRL, we show that our approach comes with a unified theory for both policy evaluation and control. Indeed, we propose two OS-DistrRL algorithms for which we provide an almost sure convergence analysis. The proposed approach compares favorably with categorical DistrRL on various environments.

Inverse reinforcement learning (IRL) methods assume that the expert data is generated by an agent optimizing some reward function. However, in many settings, the agent may optimize a reward function subject to some constraints, where the constraints induce behaviors that may be otherwise difficult to express with just a reward function. We consider the setting where the reward function is given, and the constraints are unknown, and propose a method that is able to recover these constraints satisfactorily from the expert data. While previous work has focused on recovering hard constraints, our method can recover cumulative soft constraints that the agent satisfies on average per episode. In IRL fashion, our method solves this problem by adjusting the constraint function iteratively through a constrained optimization procedure, until the agent behavior matches the expert behavior. We demonstrate our approach on synthetic environments, robotics environments and real world highway driving scenarios.

We study online learning in episodic constrained Markov decision processes (CMDPs), where the goal of the learner is to collect as much reward as possible over the episodes, while guaranteeing that some long-term constraints are satisfied during the learning process. Rewards and constraints can be selected either stochastically or adversarially, and the transition function is not known to the learner. While online learning in classical unconstrained MDPs has received considerable attention over the last years, the setting of CMDPs is still largely unexplored. This is surprising, since in real-world applications, such as, e.g., autonomous driving, automated bidding, and recommender systems, there are usually additional constraints and specifications that an agent has to obey during the learning process. In this paper, we provide the first best-of-both-worlds algorithm for CMDPs with long-term constraints. Our algorithm is capable of handling settings in which rewards and constraints are selected either stochastically or adversarially, without requiring any knowledge of the underling process. Moreover, our algorithm matches state-of-the-art regret and constraint violation bounds for settings in which constraints are selected stochastically, while it is the first to provide guarantees in the case in which they are chosen adversarially.

A crucial challenge in reinforcement learning is to reduce the number of interactions with the environment that an agent requires to master a given task. Transfer learning proposes to address this issue by re-using knowledge from previously learned tasks. However, determining which source task qualifies as the most appropriate for knowledge extraction, as well as the choice regarding which algorithm components to transfer, represent severe obstacles to its application in reinforcement learning. The goal of this paper is to address these issues with modular multi-source transfer learning techniques. The proposed techniques automatically learn how to extract useful information from source tasks, regardless of the difference in state-action space and reward function. We support our claims with extensive and challenging cross-domain experiments for visual control.