Deep reinforcement learning has emerged as a powerful tool for obtaining high-performance policies. However, the safety of these policies has been a long-standing issue. One promising paradigm to guarantee safety is a shield, which shields a policy from making unsafe actions. However, computing a shield scales exponentially in the number of state variables. This is a particular concern in multi-agent systems with many agents. In this work, we propose a novel approach for multi-agent shielding. We address scalability by computing individual shields for each agent. The challenge is that typical safety specifications are global properties, but the shields of individual agents only ensure local properties. Our key to overcome this challenge is to apply assume-guarantee reasoning. Specifically, we present a sound proof rule that decomposes a (global, complex) safety specification into (local, simple) obligations for the shields of the individual agents. Moreover, we show that applying the shields during reinforcement learning significantly improves the quality of the policies obtained for a given training budget. We demonstrate the effectiveness and scalability of our multi-agent shielding framework in two case studies, reducing the computation time from hours to seconds and achieving fast learning convergence.
Reinforcement learning has shown great promise in aligning language models with human preferences in a variety of text generation tasks, including machine translation. For translation tasks, rewards can easily be obtained from quality estimation (QE) models which can generate rewards for unlabeled data. Despite its usefulness, reinforcement learning cannot exploit the gradients with respect to the QE score. We propose QE-EBM, a method of employing quality estimators as trainable loss networks that can directly backpropagate to the NMT model. We examine our method on several low and high resource target languages with English as the source language. QE-EBM outperforms strong baselines such as REINFORCE and proximal policy optimization (PPO) as well as supervised fine-tuning for all target languages, especially low-resource target languages. Most notably, for English-to-Mongolian translation, our method achieves improvements of 2.5 BLEU, 7.1 COMET-KIWI, 5.3 COMET, and 6.4 XCOMET relative to the supervised baseline.
Deep reinforcement learning has seen successful implementations on humanoid robots to achieve dynamic walking. However, these implementations have been so far successful in simple environments void of obstacles. In this paper, we aim to achieve bipedal locomotion in an environment where obstacles are present using a policy-based reinforcement learning. By adding simple distance reward terms to a state of art reward function that can achieve basic bipedal locomotion, the trained policy succeeds in navigating the robot towards the desired destination without colliding with the obstacles along the way.
Many organisms and cell types, from bacteria to cancer cells, exhibit a remarkable ability to adapt to fluctuating environments. Additionally, cells can leverage a memory of past environments to better survive previously-encountered stressors. From a control perspective, this adaptability poses significant challenges in driving cell populations toward extinction, and thus poses an open question with great clinical significance. In this work, we focus on drug dosing in cell populations exhibiting phenotypic plasticity. For specific dynamical models switching between resistant and susceptible states, exact solutions are known. However, when the underlying system parameters are unknown, and for complex memory-based systems, obtaining the optimal solution is currently intractable. To address this challenge, we apply reinforcement learning (RL) to identify informed dosing strategies to control cell populations evolving under novel non-Markovian dynamics. We find that model-free deep RL is able to recover exact solutions and control cell populations even in the presence of long-range temporal dynamics. To further test our approach in more realistic settings, we demonstrate robust RL-based control strategies in environments with measurement noise and dynamic memory strength.
In multi-agent reinforcement learning (MARL), parameter sharing is commonly employed to enhance sample efficiency. However, the popular approach of full parameter sharing often leads to homogeneous policies among agents, potentially limiting the performance benefits that could be derived from policy diversity. To address this critical limitation, we introduce \emph{Kaleidoscope}, a novel adaptive partial parameter sharing scheme that fosters policy heterogeneity while still maintaining high sample efficiency. Specifically, Kaleidoscope maintains one set of common parameters alongside multiple sets of distinct, learnable masks for different agents, dictating the sharing of parameters. It promotes diversity among policy networks by encouraging discrepancy among these masks, without sacrificing the efficiencies of parameter sharing. This design allows Kaleidoscope to dynamically balance high sample efficiency with a broad policy representational capacity, effectively bridging the gap between full parameter sharing and non-parameter sharing across various environments. We further extend Kaleidoscope to critic ensembles in the context of actor-critic algorithms, which could help improve value estimations.Our empirical evaluations across extensive environments, including multi-agent particle environment, multi-agent MuJoCo and StarCraft multi-agent challenge v2, demonstrate the superior performance of Kaleidoscope compared with existing parameter sharing approaches, showcasing its potential for performance enhancement in MARL. The code is publicly available at \url{https://github.com/LXXXXR/Kaleidoscope}.
This paper investigates how children learn numbers using the framework of reinforcement learning (RL), with a focus on the impact of language instructions. The motivation for using reinforcement learning stems from its parallels with psychological learning theories in controlled environments. By using state of the art deep reinforcement learning models, we simulate and analyze the effects of various forms of language instructions on number acquisition. Our findings indicate that certain linguistic structures more effectively improve numerical comprehension in RL agents. Additionally, our model predicts optimal sequences for presenting numbers to RL agents which enhance their speed of learning. This research provides valuable insights into the interplay between language and numerical cognition, with implications for both educational strategies and the development of artificial intelligence systems designed to support early childhood learning.
There have been key advancements to building universal approximators for multi-goal collections of reinforcement learning value functions -- key elements in estimating long-term returns of states in a parameterized manner. We extend this to hierarchical reinforcement learning, using the options framework, by introducing hierarchical universal value function approximators (H-UVFAs). This allows us to leverage the added benefits of scaling, planning, and generalization expected in temporal abstraction settings. We develop supervised and reinforcement learning methods for learning embeddings of the states, goals, options, and actions in the two hierarchical value functions: $Q(s, g, o; \theta)$ and $Q(s, g, o, a; \theta)$. Finally we demonstrate generalization of the HUVFAs and show they outperform corresponding UVFAs.
Learning a latent dynamics model provides a task-agnostic representation of an agent's understanding of its environment. Leveraging this knowledge for model-based reinforcement learning (RL) holds the potential to improve sample efficiency over model-free methods by learning from imagined rollouts. Furthermore, because the latent space serves as input to behavior models, the informative representations learned by the world model facilitate efficient learning of desired skills. Most existing methods rely on holistic representations of the environment's state. In contrast, humans reason about objects and their interactions, predicting how actions will affect specific parts of their surroundings. Inspired by this, we propose Slot-Attention for Object-centric Latent Dynamics (SOLD), a novel model-based RL algorithm that learns object-centric dynamics models in an unsupervised manner from pixel inputs. We demonstrate that the structured latent space not only improves model interpretability but also provides a valuable input space for behavior models to reason over. Our results show that SOLD outperforms DreamerV3 and TD-MPC2 - state-of-the-art model-based RL algorithms - across a range of benchmark robotic environments that require relational reasoning and manipulation capabilities. Videos are available at https://slot-latent-dynamics.github.io/.
Cyber-attacks are becoming increasingly sophisticated and frequent, highlighting the importance of network intrusion detection systems. This paper explores the potential and challenges of using deep reinforcement learning (DRL) in network intrusion detection. It begins by introducing key DRL concepts and frameworks, such as deep Q-networks and actor-critic algorithms, and reviews recent research utilizing DRL for intrusion detection. The study evaluates challenges related to model training efficiency, detection of minority and unknown class attacks, feature selection, and handling unbalanced datasets. The performance of DRL models is comprehensively analyzed, showing that while DRL holds promise, many recent technologies remain underexplored. Some DRL models achieve state-of-the-art results on public datasets, occasionally outperforming traditional deep learning methods. The paper concludes with recommendations for enhancing DRL deployment and testing in real-world network scenarios, with a focus on Internet of Things intrusion detection. It discusses recent DRL architectures and suggests future policy functions for DRL-based intrusion detection. Finally, the paper proposes integrating DRL with generative methods to further improve performance, addressing current gaps and supporting more robust and adaptive network intrusion detection systems.
In the context of inverse reinforcement learning (IRL) with a single expert, adversarial inverse reinforcement learning (AIRL) serves as a foundational approach to providing comprehensive and transferable task descriptions. However, AIRL faces practical performance challenges, primarily stemming from the framework's overly idealized decomposability condition, the unclear proof regarding the potential equilibrium in reward recovery, or questionable robustness in high-dimensional environments. This paper revisits AIRL in \textbf{high-dimensional scenarios where the state space tends to infinity}. Specifically, we first establish a necessary and sufficient condition for reward transferability by examining the rank of the matrix derived from subtracting the identity matrix from the transition matrix. Furthermore, leveraging random matrix theory, we analyze the spectral distribution of this matrix, demonstrating that our rank criterion holds with high probability even when the transition matrices are unobservable. This suggests that the limitations on transfer are not inherent to the AIRL framework itself, but are instead related to the training variance of the reinforcement learning algorithms employed within it. Based on this insight, we propose a hybrid framework that integrates on-policy proximal policy optimization in the source environment with off-policy soft actor-critic in the target environment, leading to significant improvements in reward transfer effectiveness.
The loss of plasticity in learning agents, analogous to the solidification of neural pathways in biological brains, significantly impedes learning and adaptation in reinforcement learning due to its non-stationary nature. To address this fundamental challenge, we propose a novel approach, Neuroplastic Expansion (NE), inspired by cortical expansion in cognitive science. NE maintains learnability and adaptability throughout the entire training process by dynamically growing the network from a smaller initial size to its full dimension. Our method is designed with three key components: (1) elastic neuron generation based on potential gradients, (2) dormant neuron pruning to optimize network expressivity, and (3) neuron consolidation via experience review to strike a balance in the plasticity-stability dilemma. Extensive experiments demonstrate that NE effectively mitigates plasticity loss and outperforms state-of-the-art methods across various tasks in MuJoCo and DeepMind Control Suite environments. NE enables more adaptive learning in complex, dynamic environments, which represents a crucial step towards transitioning deep reinforcement learning from static, one-time training paradigms to more flexible, continually adapting models.
Hierarchical policies enable strong performance in many sequential decision-making problems, such as those with high-dimensional action spaces, those requiring long-horizon planning, and settings with sparse rewards. However, learning hierarchical policies from static offline datasets presents a significant challenge. Crucially, actions taken by higher-level policies may not be directly observable within hierarchical controllers, and the offline dataset might have been generated using a different policy structure, hindering the use of standard offline learning algorithms. In this work, we propose OHIO: a framework for offline reinforcement learning (RL) of hierarchical policies. Our framework leverages knowledge of the policy structure to solve the \textit{inverse problem}, recovering the unobservable high-level actions that likely generated the observed data under our hierarchical policy. This approach constructs a dataset suitable for off-the-shelf offline training. We demonstrate our framework on robotic and network optimization problems and show that it substantially outperforms end-to-end RL methods and improves robustness. We investigate a variety of instantiations of our framework, both in direct deployment of policies trained offline and when online fine-tuning is performed. Code and data are available at https://ohio-offline-hierarchical-rl.github.io
In sequential decision-making (SDM) tasks, methods like reinforcement learning (RL) and heuristic search have made notable advances in specific cases. However, they often require extensive exploration and face challenges in generalizing across diverse environments due to their limited grasp of the underlying decision dynamics. In contrast, large language models (LLMs) have recently emerged as powerful general-purpose tools, due to their capacity to maintain vast amounts of domain-specific knowledge. To harness this rich prior knowledge for efficiently solving complex SDM tasks, we propose treating LLMs as prior action distributions and integrating them into RL frameworks through Bayesian inference methods, making use of variational inference and direct posterior sampling. The proposed approaches facilitate the seamless incorporation of fixed LLM priors into both policy-based and value-based RL frameworks. Our experiments show that incorporating LLM-based action priors significantly reduces exploration and optimization complexity, substantially improving sample efficiency compared to traditional RL techniques, e.g., using LLM priors decreases the number of required samples by over 90% in offline learning scenarios.
In our recent work [3] we introduced the grid-sampling SDE as a proxy for modeling exploration in continuous-time reinforcement learning. In this note, we provide further motivation for the use of this SDE and discuss its wellposedness in the presence of jumps.
Generating human motion from textual descriptions is a challenging task. Existing methods either struggle with physical credibility or are limited by the complexities of physics simulations. In this paper, we present \emph{ReinDiffuse} that combines reinforcement learning with motion diffusion model to generate physically credible human motions that align with textual descriptions. Our method adapts Motion Diffusion Model to output a parameterized distribution of actions, making them compatible with reinforcement learning paradigms. We employ reinforcement learning with the objective of maximizing physically plausible rewards to optimize motion generation for physical fidelity. Our approach outperforms existing state-of-the-art models on two major datasets, HumanML3D and KIT-ML, achieving significant improvements in physical plausibility and motion quality. Project: https://reindiffuse.github.io/
Exploration remains a significant challenge in reinforcement learning, especially in environments where extrinsic rewards are sparse or non-existent. The recent rise of foundation models, such as CLIP, offers an opportunity to leverage pretrained, semantically rich embeddings that encapsulate broad and reusable knowledge. In this work we explore the potential of these foundation models not just to drive exploration, but also to analyze the critical role of the episodic novelty term in enhancing exploration effectiveness of the agent. We also investigate whether providing the intrinsic module with complete state information -- rather than just partial observations -- can improve exploration, despite the difficulties in handling small variations within large state spaces. Our experiments in the MiniGrid domain reveal that intrinsic modules can effectively utilize full state information, significantly increasing sample efficiency while learning an optimal policy. Moreover, we show that the embeddings provided by foundation models are sometimes even better than those constructed by the agent during training, further accelerating the learning process, especially when coupled with the episodic novelty term to enhance exploration.
Reinforcement Learning (RL) applied in healthcare can lead to unsafe medical decisions and treatment, such as excessive dosages or abrupt changes, often due to agents overlooking common-sense constraints. Consequently, Constrained Reinforcement Learning (CRL) is a natural choice for safe decisions. However, specifying the exact cost function is inherently difficult in healthcare. Recent Inverse Constrained Reinforcement Learning (ICRL) is a promising approach that infers constraints from expert demonstrations. ICRL algorithms model Markovian decisions in an interactive environment. These settings do not align with the practical requirement of a decision-making system in healthcare, where decisions rely on historical treatment recorded in an offline dataset. To tackle these issues, we propose the Constraint Transformer (CT). Specifically, 1) we utilize a causal attention mechanism to incorporate historical decisions and observations into the constraint modeling, while employing a Non-Markovian layer for weighted constraints to capture critical states. 2) A generative world model is used to perform exploratory data augmentation, enabling offline RL methods to simulate unsafe decision sequences. In multiple medical scenarios, empirical results demonstrate that CT can capture unsafe states and achieve strategies that approximate lower mortality rates, reducing the occurrence probability of unsafe behaviors.
Constrained Reinforcement Learning (CRL) is a subset of machine learning that introduces constraints into the traditional reinforcement learning (RL) framework. Unlike conventional RL which aims solely to maximize cumulative rewards, CRL incorporates additional constraints that represent specific mission requirements or limitations that the agent must comply with during the learning process. In this paper, we address a type of CRL problem where an agent aims to learn the optimal policy to maximize reward while ensuring a desired level of temporal logic constraint satisfaction throughout the learning process. We propose a novel framework that relies on switching between pure learning (reward maximization) and constraint satisfaction. This framework estimates the probability of constraint satisfaction based on earlier trials and properly adjusts the probability of switching between learning and constraint satisfaction policies. We theoretically validate the correctness of the proposed algorithm and demonstrate its performance through comprehensive simulations.
This paper showcases the use of a reinforcement learning-based Neural Architecture Search (NAS) agent to design a small neural network to perform active fire detection on multispectral satellite imagery. Specifically, we aim to design a neural network that can determine if a single multispectral pixel is a part of a fire, and do so within the constraints of a Low Earth Orbit (LEO) nanosatellite with a limited power budget, to facilitate on-board processing of sensor data. In order to use reinforcement learning, a reward function is needed. We supply this reward function in the shape of a regression model that predicts the F1 score obtained by a particular architecture, following quantization to INT8 precision, from purely architectural features. This model is trained by collecting a random sample of neural network architectures, training these architectures, and collecting their classification performance statistics. Besides the F1 score, we also include the total number of trainable parameters in our reward function to limit the size of the designed model and ensure it fits within the resource constraints imposed by nanosatellite platforms. Finally, we deployed the best neural network to the Google Coral Micro Dev Board and evaluated its inference latency and power consumption. This neural network consists of 1,716 trainable parameters, takes on average 984{\mu}s to inference, and consumes around 800mW to perform inference. These results show that our reinforcement learning-based NAS approach can be successfully applied to novel problems not tackled before.
Antibodies offer great potential for the treatment of various diseases. However, the discovery of therapeutic antibodies through traditional wet lab methods is expensive and time-consuming. The use of generative models in designing antibodies therefore holds great promise, as it can reduce the time and resources required. Recently, the class of diffusion models has gained considerable traction for their ability to synthesize diverse and high-quality samples. In their basic form, however, they lack mechanisms to optimize for specific properties, such as binding affinity to an antigen. In contrast, the class of offline Reinforcement Learning (RL) methods has demonstrated strong performance in navigating large search spaces, including scenarios where frequent real-world interaction, such as interaction with a wet lab, is impractical. Our novel method, BetterBodies, which combines Variational Autoencoders (VAEs) with RL guided latent diffusion, is able to generate novel sets of antibody CDRH3 sequences from different data distributions. Using the Absolut! simulator, we demonstrate the improved affinity of our novel sequences to the SARS-CoV spike receptor-binding domain. Furthermore, we reflect biophysical properties in the VAE latent space using a contrastive loss and add a novel Q-function based filtering to enhance the affinity of generated sequences. In conclusion, methods such as ours have the potential to have great implications for real-world biological sequence design, where the generation of novel high-affinity binders is a cost-intensive endeavor.