Replaying past experiences has proven to be a highly effective approach for averting catastrophic forgetting in supervised continual learning. However, some crucial factors are still largely ignored, making it vulnerable to serious failure, when used as a solution to forgetting in continual reinforcement learning, even in the context of perfect memory where all data of previous tasks are accessible in the current task. On the one hand, since most reinforcement learning algorithms are not invariant to the reward scale, the previously well-learned tasks (with high rewards) may appear to be more salient to the current learning process than the current task (with small initial rewards). This causes the agent to concentrate on those salient tasks at the expense of generality on the current task. On the other hand, offline learning on replayed tasks while learning a new task may induce a distributional shift between the dataset and the learned policy on old tasks, resulting in forgetting. In this paper, we introduce RECALL, a replay-enhanced method that greatly improves the plasticity of existing replay-based methods on new tasks while effectively avoiding the recurrence of catastrophic forgetting in continual reinforcement learning. RECALL leverages adaptive normalization on approximate targets and policy distillation on old tasks to enhance generality and stability, respectively. Extensive experiments on the Continual World benchmark show that RECALL performs significantly better than purely perfect memory replay, and achieves comparable or better overall performance against state-of-the-art continual learning methods.
Deep Reinforcement Learning (DRL) agents frequently face challenges in adapting to tasks outside their training distribution, including issues with over-fitting, catastrophic forgetting and sample inefficiency. Although the application of adapters has proven effective in supervised learning contexts such as natural language processing and computer vision, their potential within the DRL domain remains largely unexplored. This paper delves into the integration of adapters in reinforcement learning, presenting an innovative adaptation strategy that demonstrates enhanced training efficiency and improvement of the base-agent, experimentally in the nanoRTS environment, a real-time strategy (RTS) game simulation. Our proposed universal approach is not only compatible with pre-trained neural networks but also with rule-based agents, offering a means to integrate human expertise.
Robotic manipulation holds the potential to replace humans in the execution of tedious or dangerous tasks. However, control-based approaches are not suitable due to the difficulty of formally describing open-world manipulation in reality, and the inefficiency of existing learning methods. Thus, applying manipulation in a wide range of scenarios presents significant challenges. In this study, we propose a novel method for skill learning in robotic manipulation called Tactile Active Inference Reinforcement Learning (Tactile-AIRL), aimed at achieving efficient training. To enhance the performance of reinforcement learning (RL), we introduce active inference, which integrates model-based techniques and intrinsic curiosity into the RL process. This integration improves the algorithm's training efficiency and adaptability to sparse rewards. Additionally, we utilize a vision-based tactile sensor to provide detailed perception for manipulation tasks. Finally, we employ a model-based approach to imagine and plan appropriate actions through free energy minimization. Simulation results demonstrate that our method achieves significantly high training efficiency in non-prehensile objects pushing tasks. It enables agents to excel in both dense and sparse reward tasks with just a few interaction episodes, surpassing the SAC baseline. Furthermore, we conduct physical experiments on a gripper screwing task using our method, which showcases the algorithm's rapid learning capability and its potential for practical applications.
This paper introduces a novel data-driven approach to design a linear quadratic regulator (LQR) using a reinforcement learning (RL) algorithm that does not require a system model. The key contribution is to perform policy iteration (PI) by designing the policy evaluator as a two-layer quadratic neural network (QNN). This network is trained through convex optimization. To the best of our knowledge, this is the first time that a QNN trained through convex optimization is employed as the Q-function approximator (QFA). The main advantage is that the QNN's input-output mapping has an analytical expression as a quadratic form, which can then be used to obtain an analytical expression for policy improvement. This is in stark contrast to the available techniques in the literature that must train a second neural network to obtain policy improvement. The article establishes the convergence of the learning algorithm to the optimal control, provided the system is controllable and one starts from a stabilitzing policy. A quadrotor example demonstrates the effectiveness of the proposed approach.
Due to the empirical success of reinforcement learning, an increasing number of students study the subject. However, from our practical teaching experience, we see students entering the field (bachelor, master and early PhD) often struggle. On the one hand, textbooks and (online) lectures provide the fundamentals, but students find it hard to translate between equations and code. On the other hand, public codebases do provide practical examples, but the implemented algorithms tend to be complex, and the underlying test environments contain multiple reinforcement learning challenges at once. Although this is realistic from a research perspective, it often hinders educational conceptual understanding. To solve this issue we introduce EduGym, a set of educational reinforcement learning environments and associated interactive notebooks tailored for education. Each EduGym environment is specifically designed to illustrate a certain aspect/challenge of reinforcement learning (e.g., exploration, partial observability, stochasticity, etc.), while the associated interactive notebook explains the challenge and its possible solution approaches, connecting equations and code in a single document. An evaluation among RL students and researchers shows 86% of them think EduGym is a useful tool for reinforcement learning education. All notebooks are available from https://www.edugym.org/, while the full software package can be installed from https://github.com/RLG-Leiden/edugym.
We present a novel learning framework to solve the transport-and-packing (TAP) problem in 3D. It constitutes a full solution pipeline from partial observations of input objects via RGBD sensing and recognition to final box placement, via robotic motion planning, to arrive at a compact packing in a target container. The technical core of our method is a neural network for TAP, trained via reinforcement learning (RL), to solve the NP-hard combinatorial optimization problem. Our network simultaneously selects an object to pack and determines the final packing location, based on a judicious encoding of the continuously evolving states of partially observed source objects and available spaces in the target container, using separate encoders both enabled with attention mechanisms. The encoded feature vectors are employed to compute the matching scores and feasibility masks of different pairings of box selection and available space configuration for packing strategy optimization. Extensive experiments, including ablation studies and physical packing execution by a real robot (Universal Robot UR5e), are conducted to evaluate our method in terms of its design choices, scalability, generalizability, and comparisons to baselines, including the most recent RL-based TAP solution. We also contribute the first benchmark for TAP which covers a variety of input settings and difficulty levels.
In recent years, interest in autonomous shipping in urban waterways has increased significantly due to the trend of keeping cars and trucks out of city centers. Classical approaches such as Frenet frame based planning and potential field navigation often require tuning of many configuration parameters and sometimes even require a different configuration depending on the situation. In this paper, we propose a novel path planning approach based on reinforcement learning called Model Predictive Reinforcement Learning (MPRL). MPRL calculates a series of waypoints for the vessel to follow. The environment is represented as an occupancy grid map, allowing us to deal with any shape of waterway and any number and shape of obstacles. We demonstrate our approach on two scenarios and compare the resulting path with path planning using a Frenet frame and path planning based on a proximal policy optimization (PPO) agent. Our results show that MPRL outperforms both baselines in both test scenarios. The PPO based approach was not able to reach the goal in either scenario while the Frenet frame approach failed in the scenario consisting of a corner with obstacles. MPRL was able to safely (collision free) navigate to the goal in both of the test scenarios.
Reinforcement learning (RL) algorithms interact with their environment in a trial-and-error fashion. Such interactions can be expensive, inefficient, and timely when learning on a physical system rather than in a simulation. This work develops new runtime verification techniques to predict when the learning phase has not met or will not meet qualitative and timely expectations. This paper presents three verification properties concerning the quality and timeliness of learning in RL algorithms. With each property, we propose design steps for monitoring and assessing the properties during the system's operation.
In addressing control problems such as regulation and tracking through reinforcement learning, it is often required to guarantee that the acquired policy meets essential performance and stability criteria such as a desired settling time and steady-state error prior to deployment. Motivated by this necessity, we present a set of results and a systematic reward shaping procedure that (i) ensures the optimal policy generates trajectories that align with specified control requirements and (ii) allows to assess whether any given policy satisfies them. We validate our approach through comprehensive numerical experiments conducted in two representative environments from OpenAI Gym: the Inverted Pendulum swing-up problem and the Lunar Lander. Utilizing both tabular and deep reinforcement learning methods, our experiments consistently affirm the efficacy of our proposed framework, highlighting its effectiveness in ensuring policy adherence to the prescribed control requirements.
Humans possess the ability to draw on past experiences explicitly when learning new tasks and applying them accordingly. We believe this capacity for self-referencing is especially advantageous for reinforcement learning agents in the unsupervised pretrain-then-finetune setting. During pretraining, an agent's past experiences can be explicitly utilized to mitigate the nonstationarity of intrinsic rewards. In the finetuning phase, referencing historical trajectories prevents the unlearning of valuable exploratory behaviors. Motivated by these benefits, we propose the Self-Reference (SR) approach, an add-on module explicitly designed to leverage historical information and enhance agent performance within the pretrain-finetune paradigm. Our approach achieves state-of-the-art results in terms of Interquartile Mean (IQM) performance and Optimality Gap reduction on the Unsupervised Reinforcement Learning Benchmark for model-free methods, recording an 86% IQM and a 16% Optimality Gap. Additionally, it improves current algorithms by up to 17% IQM and reduces the Optimality Gap by 31%. Beyond performance enhancement, the Self-Reference add-on also increases sample efficiency, a crucial attribute for real-world applications.
Utilizing Deep Reinforcement Learning (DRL) for Reconfigurable Intelligent Surface (RIS) assisted wireless communication has been extensively researched. However, existing DRL methods either act as a simple optimizer or only solve problems with concurrent Channel State Information (CSI) represented in the training data set. Consequently, solutions for RIS-assisted wireless communication systems under time-varying environments are relatively unexplored. However, communication problems should be considered with realistic assumptions; for instance, in scenarios where the channel is time-varying, the policy obtained by reinforcement learning should be applicable for situations where CSI is not well represented in the training data set. In this paper, we apply Meta-Reinforcement Learning (MRL) to the joint optimization problem of active beamforming at the Base Station (BS) and phase shift at the RIS, motivated by MRL's ability to extend the DRL concept of solving one Markov Decision Problem (MDP) to multiple MDPs. We provide simulation results to compare the average sum rate of the proposed approach with those of selected forerunners in the literature. Our approach improves the sum rate by more than 60% under time-varying CSI assumption while maintaining the advantages of typical DRL-based solutions. Our study's results emphasize the possibility of utilizing MRL-based designs in RIS-assisted wireless communication systems while considering realistic environment assumptions.
A significant bottleneck in applying current reinforcement learning algorithms to real-world scenarios is the need to reset the environment between every episode. This reset process demands substantial human intervention, making it difficult for the agent to learn continuously and autonomously. Several recent works have introduced autonomous reinforcement learning (ARL) algorithms that generate curricula for jointly training reset and forward policies. While their curricula can reduce the number of required manual resets by taking into account the agent's learning progress, they rely on task-specific knowledge, such as predefined initial states or reset reward functions. In this paper, we propose a novel ARL algorithm that can generate a curriculum adaptive to the agent's learning progress without task-specific knowledge. Our curriculum empowers the agent to autonomously reset to diverse and informative initial states. To achieve this, we introduce a success discriminator that estimates the success probability from each initial state when the agent follows the forward policy. The success discriminator is trained with relabeled transitions in a self-supervised manner. Our experimental results demonstrate that our ARL algorithm can generate an adaptive curriculum and enable the agent to efficiently bootstrap to solve sparse-reward maze navigation and manipulation tasks, outperforming baselines with significantly fewer manual resets.
Motivated by the need for a robust policy in the face of environment shifts between training and the deployment, we contribute to the theoretical foundation of distributionally robust reinforcement learning (DRRL). This is accomplished through a comprehensive modeling framework centered around distributionally robust Markov decision processes (DRMDPs). This framework obliges the decision maker to choose an optimal policy under the worst-case distributional shift orchestrated by an adversary. By unifying and extending existing formulations, we rigorously construct DRMDPs that embraces various modeling attributes for both the decision maker and the adversary. These attributes include adaptability granularity, exploring history-dependent, Markov, and Markov time-homogeneous decision maker and adversary dynamics. Additionally, we delve into the flexibility of shifts induced by the adversary, examining SA and S-rectangularity. Within this DRMDP framework, we investigate conditions for the existence or absence of the dynamic programming principle (DPP). From an algorithmic standpoint, the existence of DPP holds significant implications, as the vast majority of existing data and computationally efficiency RL algorithms are reliant on the DPP. To study its existence, we comprehensively examine combinations of controller and adversary attributes, providing streamlined proofs grounded in a unified methodology. We also offer counterexamples for settings in which a DPP with full generality is absent.
Robustness to noise is of utmost importance in reinforcement learning systems, particularly in military contexts where high stakes and uncertain environments prevail. Noise and uncertainty are inherent features of military operations, arising from factors such as incomplete information, adversarial actions, or unpredictable battlefield conditions. In RL, noise can critically impact decision-making, mission success, and the safety of personnel. Reward machines offer a powerful tool to express complex reward structures in RL tasks, enabling the design of tailored reinforcement signals that align with mission objectives. This paper considers the problem of the robustness of intelligence-driven reinforcement learning based on reward machines. The preliminary results presented suggest the need for further research in evidential reasoning and learning to harden current state-of-the-art reinforcement learning approaches before being mission-critical-ready.
In recent years, Reward Machines (RMs) have stood out as a simple yet effective automata-based formalism for exposing and exploiting task structure in reinforcement learning settings. Despite their relevance, little to no attention has been directed to the study of their security implications and robustness to adversarial scenarios, likely due to their recent appearance in the literature. With my thesis, I aim to provide the first analysis of the security of RM-based reinforcement learning techniques, with the hope of motivating further research in the field, and I propose and evaluate a novel class of attacks on RM-based techniques: blinding attacks.
Offline reinforcement learning suffers from the out-of-distribution issue and extrapolation error. Most policy constraint methods regularize the density of the trained policy towards the behavior policy, which is too restrictive in most cases. We propose Supported Trust Region optimization (STR) which performs trust region policy optimization with the policy constrained within the support of the behavior policy, enjoying the less restrictive support constraint. We show that, when assuming no approximation and sampling error, STR guarantees strict policy improvement until convergence to the optimal support-constrained policy in the dataset. Further with both errors incorporated, STR still guarantees safe policy improvement for each step. Empirical results validate the theory of STR and demonstrate its state-of-the-art performance on MuJoCo locomotion domains and much more challenging AntMaze domains.
In the backdrop of an increasingly pressing need for effective urban and highway transportation systems, this work explores the synergy between model-based and learning-based strategies to enhance traffic flow management by use of an innovative approach to the problem of ramp metering control that embeds Reinforcement Learning (RL) techniques within the Model Predictive Control (MPC) framework. The control problem is formulated as an RL task by crafting a suitable stage cost function that is representative of the traffic conditions, variability in the control action, and violations of the constraint on the maximum number of vehicles in queue. An MPC-based RL approach, which leverages the MPC optimal problem as a function approximation for the RL algorithm, is proposed to learn to efficiently control an on-ramp and satisfy its constraints despite uncertainties in the system model and variable demands. Simulations are performed on a benchmark small-scale highway network to compare the proposed methodology against other state-of-the-art control approaches. Results show that, starting from an MPC controller that has an imprecise model and is poorly tuned, the proposed methodology is able to effectively learn to improve the control policy such that congestion in the network is reduced and constraints are satisfied, yielding an improved performance that is superior to the other controllers.
Inverse Reinforcement Learning (IRL) learns an optimal policy, given some expert demonstrations, thus avoiding the need for the tedious process of specifying a suitable reward function. However, current methods are constrained by at least one of the following requirements. The first one is the need to fully solve a forward Reinforcement Learning (RL) problem in the inner loop of the algorithm, which might be prohibitively expensive in many complex environments. The second one is the need for full trajectories from the experts, which might not be easily available. The third one is the assumption that the expert data is homogeneous rather than a collection from various experts or possibly alternative solutions to the same task. Such constraints make IRL approaches either not scalable or not usable on certain existing systems. In this work we propose an approach which removes these requirements through a dynamic, adaptive method called Adversarial Imitation Learning on Aggregated Data (AILAD). It learns conjointly both a non linear reward function and the associated optimal policy using an adversarial framework. The reward learner only uses aggregated data. Moreover, it generates diverse behaviors producing a distribution over the aggregated data matching that of the experts.
Minimum Bayes Risk (MBR) decoding can significantly improve translation performance of Multilingual Large Language Models (MLLMs). However, MBR decoding is computationally expensive. We show how the recently developed Reinforcement Learning technique, Direct Preference Optimization (DPO), can fine-tune MLLMs to get the gains of MBR without any additional computation in inference. Our method uses only a small monolingual fine-tuning set and yields significantly improved performance on multiple NMT test sets compared to MLLMs without DPO.
On-policy reinforcement learning (RL) algorithms perform policy updates using i.i.d. trajectories collected by the current policy. However, after observing only a finite number of trajectories, on-policy sampling may produce data that fails to match the expected on-policy data distribution. This sampling error leads to noisy updates and data inefficient on-policy learning. Recent work in the policy evaluation setting has shown that non-i.i.d., off-policy sampling can produce data with lower sampling error than on-policy sampling can produce (Zhong et. al, 2022). Motivated by this observation, we introduce an adaptive, off-policy sampling method to improve the data efficiency of on-policy policy gradient algorithms. Our method, Proximal Robust On-Policy Sampling (PROPS), reduces sampling error by collecting data with a behavior policy that increases the probability of sampling actions that are under-sampled with respect to the current policy. We empirically evaluate PROPS on both continuous-action MuJoCo benchmark tasks as well discrete-action tasks and demonstrate that (1) PROPS decreases sampling error throughout training and (2) improves the data efficiency of on-policy policy gradient algorithms.