Robust policies enable reinforcement learning agents to effectively adapt to and operate in unpredictable, dynamic, and ever-changing real-world environments. Factored representations, which break down complex state and action spaces into distinct components, can improve generalization and sample efficiency in policy learning. In this paper, we explore how the curriculum of an agent using a factored state representation affects the robustness of the learned policy. We experimentally demonstrate three simple curricula, such as varying only the variable of highest regret between episodes, that can significantly enhance policy robustness, offering practical insights for reinforcement learning in complex environments.
This work presents and showcases a novel reinforcement learning agent called Critic As Lyapunov Function (CALF) which is model-free and ensures online environment, in other words, dynamical system stabilization. Online means that in each learning episode, the said environment is stabilized. This, as demonstrated in a case study with a mobile robot simulator, greatly improves the overall learning performance. The base actor-critic scheme of CALF is analogous to SARSA. The latter did not show any success in reaching the target in our studies. However, a modified version thereof, called SARSA-m here, did succeed in some learning scenarios. Still, CALF greatly outperformed the said approach. CALF was also demonstrated to improve a nominal stabilizer provided to it. In summary, the presented agent may be considered a viable approach to fusing classical control with reinforcement learning. Its concurrent approaches are mostly either offline or model-based, like, for instance, those that fuse model-predictive control into the agent.
Kolmogorov-Arnold Networks (KAN) is an emerging neural network architecture in machine learning. It has greatly interested the research community about whether KAN can be a promising alternative of the commonly used Multi-Layer Perceptions (MLP). Experiments in various fields demonstrated that KAN-based machine learning can achieve comparable if not better performance than MLP-based methods, but with much smaller parameter scales and are more explainable. In this paper, we explore the incorporation of KAN into the actor and critic networks for offline reinforcement learning (RL). We evaluated the performance, parameter scales, and training efficiency of various KAN and MLP based conservative Q-learning (CQL) on the the classical D4RL benchmark for offline RL. Our study demonstrates that KAN can achieve performance close to the commonly used MLP with significantly fewer parameters. This provides us an option to choose the base networks according to the requirements of the offline RL tasks.
Supervised learning approaches to offline reinforcement learning, particularly those utilizing the Decision Transformer, have shown effectiveness in continuous environments and for sparse rewards. However, they often struggle with long-horizon tasks due to the high compounding error of auto-regressive models. To overcome this limitation, we go beyond next-token prediction and introduce Planning Tokens, which contain high-level, long time-scale information about the agent's future. Predicting dual time-scale tokens at regular intervals enables our model to use these long-horizon Planning Tokens as a form of implicit planning to guide its low-level policy and reduce compounding error. This architectural modification significantly enhances performance on long-horizon tasks, establishing a new state-of-the-art in complex D4RL environments. Additionally, we demonstrate that Planning Tokens improve the interpretability of the model's policy through the interpretable plan visualisations and attention map.
Reinforcement learning from human feedback (RLHF) has become a key method for aligning large language models (LLMs) with human preferences through the use of reward models. However, traditional reward models typically generate point estimates, which oversimplify the diversity and complexity of human values and preferences. In this paper, we introduce Quantile Reward Models (QRMs), a novel approach to reward modeling that learns a distribution over rewards instead of a single scalar value. Our method uses quantile regression to estimate a full, potentially multimodal distribution over preferences, providing a more powerful and nuanced representation of preferences. This distributional approach can better capture the diversity of human values, addresses label noise, and accommodates conflicting preferences by modeling them as distinct modes in the distribution. Our experimental results show that QRM outperforms comparable traditional point-estimate models on RewardBench. Furthermore, we demonstrate that the additional information provided by the distributional estimates can be utilized in downstream applications, such as risk-aware reinforcement learning, resulting in LLM policies that generate fewer extremely negative responses. Our code and model are released at https://github.com/Nicolinho/QRM.
Although deep reinforcement learning (DRL) approaches in audio signal processing have seen substantial progress in recent years, audio-driven DRL for tasks such as navigation, gaze control and head-orientation control in the context of human-robot interaction have received little attention. Here, we propose an audio-driven DRL framework in which we utilise deep Q-learning to develop an autonomous agent that orients towards a talker in the acoustic environment based on stereo speech recordings. Our results show that the agent learned to perform the task at a near perfect level when trained on speech segments in anechoic environments (that is, without reverberation). The presence of reverberation in naturalistic acoustic environments affected the agent's performance, although the agent still substantially outperformed a baseline, randomly acting agent. Finally, we quantified the degree of generalization of the proposed DRL approach across naturalistic acoustic environments. Our experiments revealed that policies learned by agents trained on medium or high reverb environments generalized to low reverb environments, but policies learned by agents trained on anechoic or low reverb environments did not generalize to medium or high reverb environments. Taken together, this study demonstrates the potential of audio-driven DRL for tasks such as head-orientation control and highlights the need for training strategies that enable robust generalization across environments for real-world audio-driven DRL applications.
Designing a stabilizing controller for nonlinear systems is a challenging task, especially for high-dimensional problems with unknown dynamics. Traditional reinforcement learning algorithms applied to stabilization tasks tend to drive the system close to the equilibrium point. However, these approaches often fall short of achieving true stabilization and result in persistent oscillations around the equilibrium point. In this work, we propose a reinforcement learning algorithm that stabilizes the system by learning a local linear representation ofthe dynamics. The main component of the algorithm is integrating the learned gain matrix directly into the neural policy. We demonstrate the effectiveness of our algorithm on several challenging high-dimensional dynamical systems. In these simulations, our algorithm outperforms popular reinforcement learning algorithms, such as soft actor-critic (SAC) and proximal policy optimization (PPO), and successfully stabilizes the system. To support the numerical results, we provide a theoretical analysis of the feasibility of the learned algorithm for both deterministic and stochastic reinforcement learning settings, along with a convergence analysis of the proposed learning algorithm. Furthermore, we verify that the learned control policies indeed provide asymptotic stability for the nonlinear systems.
Reinforcement Learning from human feedback (RLHF) has been shown a promising direction for aligning generative models with human intent and has also been explored in recent works for alignment of diffusion generative models. In this work, we provide a rigorous treatment by formulating the task of fine-tuning diffusion models, with reward functions learned from human feedback, as an exploratory continuous-time stochastic control problem. Our key idea lies in treating the score-matching functions as controls/actions, and upon this, we develop a unified framework from a continuous-time perspective, to employ reinforcement learning (RL) algorithms in terms of improving the generation quality of diffusion models. We also develop the corresponding continuous-time RL theory for policy optimization and regularization under assumptions of stochastic different equations driven environment. Experiments on the text-to-image (T2I) generation will be reported in the accompanied paper.
We have developed a reinforcement learning agent that often finds a minimal sequence of unknotting crossing changes for a knot diagram with up to 200 crossings, hence giving an upper bound on the unknotting number. We have used this to determine the unknotting number of 57k knots. We took diagrams of connected sums of such knots with oppositely signed signatures, where the summands were overlaid. The agent has found examples where several of the crossing changes in an unknotting collection of crossings result in hyperbolic knots. Based on this, we have shown that, given knots $K$ and $K'$ that satisfy some mild assumptions, there is a diagram of their connected sum and $u(K) + u(K')$ unknotting crossings such that changing any one of them results in a prime knot. As a by-product, we have obtained a dataset of 2.6 million distinct hard unknot diagrams; most of them under 35 crossings. Assuming the additivity of the unknotting number, we have determined the unknotting number of 43 at most 12-crossing knots for which the unknotting number is unknown.
Optical wireless communication (OWC) is envisioned as one of the main enabling technologies of 6G networks, complementing radio frequency (RF) systems to provide high data rates. One of the crucial issues in indoor OWC is service interruptions due to blockages that obstruct the line of sight (LoS) between users and their access points (APs). Recently, reflecting surfaces referred to as intelligent reflecting surfaces (IRSs) have been considered to provide improved connectivity in OWC systems by reflecting AP signals toward users. In this study, we investigate the integration of IRSs into an indoor OWC system to improve the sum rate of the users and to ensure service continuity. We formulate an optimization problem for sum rate maximization, where the allocation of both APs and mirror elements of IRSs to users is determined to enhance the aggregate data rate. Moreover, reinforcement learning (RL) algorithms, specifically Q-learning and SARSA algorithms, are proposed to provide real-time solutions with low complexity and without prior system knowledge. The results show that the RL algorithms achieve near-optimal solutions that are close to the solutions of mixed integer linear programming (MILP). The results also show that the proposed scheme achieves up to a 45% increase in data rate compared to a traditional scheme that optimizes only the allocation of APs while the mirror elements are assigned to users based on the distance.
The increase of renewable energy generation towards the zero-emission target is making the problem of controlling power grids more and more challenging. The recent series of competitions Learning To Run a Power Network (L2RPN) have encouraged the use of Reinforcement Learning (RL) for the assistance of human dispatchers in operating power grids. All the solutions proposed so far severely restrict the action space and are based on a single agent acting on the entire grid or multiple independent agents acting at the substations level. In this work, we propose a domain-agnostic algorithm that estimates correlations between state and action components entirely based on data. Highly correlated state-action pairs are grouped together to create simpler, possibly independent subproblems that can lead to distinct learning processes with less computational and data requirements. The algorithm is validated on a power grid benchmark obtained with the Grid2Op simulator that has been used throughout the aforementioned competitions, showing that our algorithm is in line with domain-expert analysis. Based on these results, we lay a theoretically-grounded foundation for using distributed reinforcement learning in order to improve the existing solutions.
As a data-driven paradigm, offline reinforcement learning (Offline RL) has been formulated as sequence modeling, where the Decision Transformer (DT) has demonstrated exceptional capabilities. Unlike previous reinforcement learning methods that fit value functions or compute policy gradients, DT adjusts the autoregressive model based on the expected returns, past states, and actions, using a causally masked Transformer to output the optimal action. However, due to the inconsistency between the sampled returns within a single trajectory and the optimal returns across multiple trajectories, it is challenging to set an expected return to output the optimal action and stitch together suboptimal trajectories. Decision ConvFormer (DC) is easier to understand in the context of modeling RL trajectories within a Markov Decision Process compared to DT. We propose the Q-value Regularized Decision ConvFormer (QDC), which combines the understanding of RL trajectories by DC and incorporates a term that maximizes action values using dynamic programming methods during training. This ensures that the expected returns of the sampled actions are consistent with the optimal returns. QDC achieves excellent performance on the D4RL benchmark, outperforming or approaching the optimal level in all tested environments. It particularly demonstrates outstanding competitiveness in trajectory stitching capability.
Inverse reinforcement learning methods aim to retrieve the reward function of a Markov decision process based on a dataset of expert demonstrations. The commonplace scarcity and heterogeneous sources of such demonstrations can lead to the absorption of spurious correlations in the data by the learned reward function. Consequently, this adaptation often exhibits behavioural overfitting to the expert data set when a policy is trained on the obtained reward function under distribution shift of the environment dynamics. In this work, we explore a novel regularization approach for inverse reinforcement learning methods based on the causal invariance principle with the goal of improved reward function generalization. By applying this regularization to both exact and approximate formulations of the learning task, we demonstrate superior policy performance when trained using the recovered reward functions in a transfer setting
Graph path search is a classic computer science problem that has been recently approached with Reinforcement Learning (RL) due to its potential to outperform prior methods. Existing RL techniques typically assume a global view of the network, which is not suitable for large-scale, dynamic, and privacy-sensitive settings. An area of particular interest is search in social networks due to its numerous applications. Inspired by seminal work in experimental sociology, which showed that decentralized yet efficient search is possible in social networks, we frame the problem as a collaborative task between multiple agents equipped with a limited local view of the network. We propose a multi-agent approach for graph path search that successfully leverages both homophily and structural heterogeneity. Our experiments, carried out over synthetic and real-world social networks, demonstrate that our model significantly outperforms learned and heuristic baselines. Furthermore, our results show that meaningful embeddings for graph navigation can be constructed using reward-driven learning.
This paper presents Tidal-MerzA, a novel system designed for collaborative performances between humans and a machine agent in the context of live coding, specifically focusing on the generation of musical patterns. Tidal-MerzA fuses two foundational models: ALCAA (Affective Live Coding Autonomous Agent) and Tidal Fuzz, a computational framework. By integrating affective modelling with computational generation, this system leverages reinforcement learning techniques to dynamically adapt music composition parameters within the TidalCycles framework, ensuring both affective qualities to the patterns and syntactical correctness. The development of Tidal-MerzA introduces two distinct agents: one focusing on the generation of mini-notation strings for musical expression, and another on the alignment of music with targeted affective states through reinforcement learning. This approach enhances the adaptability and creative potential of live coding practices and allows exploration of human-machine creative interactions. Tidal-MerzA advances the field of computational music generation, presenting a novel methodology for incorporating artificial intelligence into artistic practices.
It has previously been shown that by using reinforcement learning (RL), agents can derive simple approximate and exact-restricted numeral systems that are similar to human ones (Carlsson, 2021). However, it is a major challenge to show how more complex recursive numeral systems, similar to for example English, could arise via a simple learning mechanism such as RL. Here, we introduce an approach towards deriving a mechanistic explanation of the emergence of efficient recursive number systems. We consider pairs of agents learning how to communicate about numerical quantities through a meta-grammar that can be gradually modified throughout the interactions. Utilising a slightly modified version of the meta-grammar of Hurford (1975), we demonstrate that our RL agents, shaped by the pressures for efficient communication, can effectively modify their lexicon towards Pareto-optimal configurations which are comparable to those observed within human numeral systems in terms of their efficiency.
Modern listwise recommendation systems need to consider both long-term user perceptions and short-term interest shifts. Reinforcement learning can be applied on recommendation to study such a problem but is also subject to large search space, sparse user feedback and long interactive latency. Motivated by recent progress in hierarchical reinforcement learning, we propose a novel framework called mccHRL to provide different levels of temporal abstraction on listwise recommendation. Within the hierarchical framework, the high-level agent studies the evolution of user perception, while the low-level agent produces the item selection policy by modeling the process as a sequential decision-making problem. We argue that such framework has a well-defined decomposition of the outra-session context and the intra-session context, which are encoded by the high-level and low-level agents, respectively. To verify this argument, we implement both a simulator-based environment and an industrial dataset-based experiment. Results observe significant performance improvement by our method, compared with several well-known baselines. Data and codes have been made public.
Deep Reinforcement Learning techniques are achieving state-of-the-art results in robust legged locomotion. While there exists a wide variety of legged platforms such as quadruped, humanoids, and hexapods, the field is still missing a single learning framework that can control all these different embodiments easily and effectively and possibly transfer, zero or few-shot, to unseen robot embodiments. We introduce URMA, the Unified Robot Morphology Architecture, to close this gap. Our framework brings the end-to-end Multi-Task Reinforcement Learning approach to the realm of legged robots, enabling the learned policy to control any type of robot morphology. The key idea of our method is to allow the network to learn an abstract locomotion controller that can be seamlessly shared between embodiments thanks to our morphology-agnostic encoders and decoders. This flexible architecture can be seen as a potential first step in building a foundation model for legged robot locomotion. Our experiments show that URMA can learn a locomotion policy on multiple embodiments that can be easily transferred to unseen robot platforms in simulation and the real world.
Real-world problems often involve complex objective structures that resist distillation into reinforcement learning environments with a single objective. Operation costs must be balanced with multi-dimensional task performance and end-states' effects on future availability, all while ensuring safety for other agents in the environment and the reinforcement learning agent itself. System redundancy through secondary backup controllers has proven to be an effective method to ensure safety in real-world applications where the risk of violating constraints is extremely high. In this work, we investigate the utility of a stacked, continuous-control variation of universal successor feature approximation (USFA) adapted for soft actor-critic (SAC) and coupled with a suite of secondary safety controllers, which we call stacked USFA for safety (SUSFAS). Our method improves performance on secondary objectives compared to SAC baselines using an intervening secondary controller such as a runtime assurance (RTA) controller.
Variational Quantum Algorithms have emerged as promising tools for solving optimization problems on quantum computers. These algorithms leverage a parametric quantum circuit called ansatz, where its parameters are adjusted by a classical optimizer with the goal of optimizing a certain cost function. However, a significant challenge lies in designing effective circuits for addressing specific problems. In this study, we leverage the powerful and flexible Reinforcement Learning paradigm to train an agent capable of autonomously generating quantum circuits that can be used as ansatzes in variational algorithms to solve optimization problems. The agent is trained on diverse problem instances, including Maximum Cut, Maximum Clique and Minimum Vertex Cover, built from different graph topologies and sizes. Our analysis of the circuits generated by the agent and the corresponding solutions shows that the proposed method is able to generate effective ansatzes. While our goal is not to propose any new specific ansatz, we observe how the agent has discovered a novel family of ansatzes effective for Maximum Cut problems, which we call $R_{yz}$-connected. We study the characteristics of one of these ansatzes by comparing it against state-of-the-art quantum algorithms across instances of varying graph topologies, sizes, and problem types. Our results indicate that the $R_{yz}$-connected circuit achieves high approximation ratios for Maximum Cut problems, further validating our proposed agent. In conclusion, our study highlights the potential of Reinforcement Learning techniques in assisting researchers to design effective quantum circuits which could have applications in a wide number of tasks.