It is quite challenging to ensure the safety of reinforcement learning (RL) agents in an unknown and stochastic environment under hard constraints that require the system state not to reach certain specified unsafe regions. Many popular safe RL methods such as those based on the Constrained Markov Decision Process (CMDP) paradigm formulate safety violations in a cost function and try to constrain the expectation of cumulative cost under a threshold. However, it is often difficult to effectively capture and enforce hard reachability-based safety constraints indirectly with such constraints on safety violation costs. In this work, we leverage the notion of barrier function to explicitly encode the hard safety constraints, and given that the environment is unknown, relax them to our design of \emph{generative-model-based soft barrier functions}. Based on such soft barriers, we propose a safe RL approach that can jointly learn the environment and optimize the control policy, while effectively avoiding unsafe regions with safety probability optimization. Experiments on a set of examples demonstrate that our approach can effectively enforce hard safety constraints and significantly outperform CMDP-based baseline methods in system safe rate measured via simulations.

Proteolysis-Targeting Chimeras (PROTACs) represent a novel class of small molecules which are designed to act as a bridge between an E3 ligase and a disease-relevant protein, thereby promoting its subsequent degradation. PROTACs are composed of two protein binding "active" domains, linked by a "linker" domain. The design of the linker domain is challenging due to geometric and chemical constraints given by its interactions, and the need to maximize drug-likeness. To tackle these challenges, we introduce ShapeLinker, a method for de novo design of linkers. It performs fragment-linking using reinforcement learning on an autoregressive SMILES generator. The method optimizes for a composite score combining relevant physicochemical properties and a novel, attention-based point cloud alignment score. This new method successfully generates linkers that satisfy both relevant 2D and 3D requirements, and achieves state-of-the-art results in producing novel linkers assuming a target linker conformation. This allows for more rational and efficient PROTAC design and optimization. Code and data are available at https://github.com/aivant/ShapeLinker.

This paper introduces a deep reinforcement learning (RL) framework for optimizing the operations of power plants pairing renewable energy with storage. The objective is to maximize revenue from energy markets while minimizing storage degradation costs and renewable curtailment. The framework handles complexities such as time coupling by storage devices, uncertainty in renewable generation and energy prices, and non-linear storage models. The study treats the problem as a hierarchical Markov Decision Process (MDP) and uses component-level simulators for storage. It utilizes RL to incorporate complex storage models, overcoming restrictions of optimization-based methods that require convex and differentiable component models. A significant aspect of this approach is ensuring policy actions respect system constraints, achieved via a novel method of projecting potentially infeasible actions onto a safe state-action set. The paper demonstrates the efficacy of this approach through extensive experiments using data from US and Indian electricity markets, comparing the learned RL policies with a baseline control policy and a retrospective optimal control policy. It validates the adaptability of the learning framework with various storage models and shows the effectiveness of RL in a complex energy optimization setting, in the context of multi-market bidding, probabilistic forecasts, and accurate storage component models.

We characterize offline data poisoning attacks on Multi-Agent Reinforcement Learning (MARL), where an attacker may change a data set in an attempt to install a (potentially fictitious) unique Markov-perfect Nash equilibrium. We propose the unique Nash set, namely the set of games, specified by their Q functions, with a specific joint policy being the unique Nash equilibrium. The unique Nash set is central to poisoning attacks because the attack is successful if and only if data poisoning pushes all plausible games inside it. The unique Nash set generalizes the reward polytope commonly used in inverse reinforcement learning to MARL. For zero-sum Markov games, both the inverse Nash set and the set of plausible games induced by data are polytopes in the Q function space. We exhibit a linear program to efficiently compute the optimal poisoning attack. Our work sheds light on the structure of data poisoning attacks on offline MARL, a necessary step before one can design more robust MARL algorithms.

Deep reinforcement learning algorithms typically act on the same set of actions. However, this is not sufficient for a wide range of real-world applications where different subsets are available at each step. In this thesis, we consider the problem of interval restrictions as they occur in pathfinding with dynamic obstacles. When actions that lead to collisions are avoided, the continuous action space is split into variable parts. Recent research learns with strong assumptions on the number of intervals, is limited to convex subsets, and the available actions are learned from the observations. Therefore, we propose two approaches that are independent of the state of the environment by extending parameterized reinforcement learning and ConstraintNet to handle an arbitrary number of intervals. We demonstrate their performance in an obstacle avoidance task and compare the methods to penalties, projection, replacement, as well as discrete and continuous masking from the literature. The results suggest that discrete masking of action-values is the only effective method when constraints did not emerge during training. When restrictions are learned, the decision between projection, masking, and our ConstraintNet modification seems to depend on the task at hand. We compare the results with varying complexity and give directions for future work.

Current reinforcement learning policies are heavily dependent upon the usage of a reward function. However, for advanced problems such as learning behavior for a self-driving car or a robot with many degrees of freedom, the reward function is very difficult or impossible to design to effectively encompass all human considerations. Inverse reinforcement learning algorithms are a class of algorithms that attempt to solve this issue by learning a reward function from expert demonstrations, and then subsequently using any manner of standard reinforcement learning algorithms to learn an optimal policy based off that learned reward function [1]. Inverse reinforcement learning algorithms require a dataset of expert demonstrations to derive a reward function from. However, this reward function can often be inaccurate and highly variable.

This paper tackles the fundamental passive eavesdropping problem in modern wireless communications in which the location and the channel state information (CSI) of the attackers are unknown. In this regard, we propose deploying an unmanned aerial vehicle (UAV) that serves as a mobile aerial relay (AR) to help ground base station (GBS) support a subset of vulnerable users. More precisely, our solution (1) clusters the single-antenna users in two groups to be either served by the GBS directly or via the AR, (2) employs optimal multi-user beamforming to the directly served users, and (3) optimizes the AR's 3D position, its multi-user beamforming matrix and transmit powers by combining closed-form solutions with machine learning techniques. Specifically, we design a plain beamforming and power optimization combined with a deep reinforcement learning (DRL) algorithm for an AR to optimize its trajectory for the security maximization of the served users. Numerical results show that the multi-user multiple input, single output (MU-MISO) system split between a GBS and an AR with optimized transmission parameters without knowledge of the eavesdropping channels achieves high secrecy capacities that scale well with increasing the number of users.

In the field of reinforcement learning (RL), agents are often tasked with solving a variety of problems differing only in their reward functions. In order to quickly obtain solutions to unseen problems with new reward functions, a popular approach involves functional composition of previously solved tasks. However, previous work using such functional composition has primarily focused on specific instances of composition functions whose limiting assumptions allow for exact zero-shot composition. Our work unifies these examples and provides a more general framework for compositionality in both standard and entropy-regularized RL. We find that, for a broad class of functions, the optimal solution for the composite task of interest can be related to the known primitive task solutions. Specifically, we present double-sided inequalities relating the optimal composite value function to the value functions for the primitive tasks. We also show that the regret of using a zero-shot policy can be bounded for this class of functions. The derived bounds can be used to develop clipping approaches for reducing uncertainty during training, allowing agents to quickly adapt to new tasks.

Continual reinforcement learning aims to sequentially learn a variety of tasks, retaining the ability to perform previously encountered tasks while simultaneously developing new policies for novel tasks. However, current continual RL approaches overlook the fact that certain tasks are identical under basic group operations like rotations or translations, especially with visual inputs. They may unnecessarily learn and maintain a new policy for each similar task, leading to poor sample efficiency and weak generalization capability. To address this, we introduce a unique Continual Vision-based Reinforcement Learning method that recognizes Group Symmetries, called COVERS, cultivating a policy for each group of equivalent tasks rather than individual tasks. COVERS employs a proximal policy optimization-based RL algorithm with an equivariant feature extractor and a novel task grouping mechanism that relies on the extracted invariant features. We evaluate COVERS on sequences of table-top manipulation tasks that incorporate image observations and robot proprioceptive information in both simulations and on real robot platforms. Our results show that COVERS accurately assigns tasks to their respective groups and significantly outperforms existing methods in terms of generalization capability.

Imitation learning has achieved great success in many sequential decision-making tasks, in which a neural agent is learned by imitating collected human demonstrations. However, existing algorithms typically require a large number of high-quality demonstrations that are difficult and expensive to collect. Usually, a trade-off needs to be made between demonstration quality and quantity in practice. Targeting this problem, in this work we consider the imitation of sub-optimal demonstrations, with both a small clean demonstration set and a large noisy set. Some pioneering works have been proposed, but they suffer from many limitations, e.g., assuming a demonstration to be of the same optimality throughout time steps and failing to provide any interpretation w.r.t knowledge learned from the noisy set. Addressing these problems, we propose {\method} by evaluating and imitating at the sub-demonstration level, encoding action primitives of varying quality into different skills. Concretely, {\method} consists of a high-level controller to discover skills and a skill-conditioned module to capture action-taking policies, and is trained following a two-phase pipeline by first discovering skills with all demonstrations and then adapting the controller to only the clean set. A mutual-information-based regularization and a dynamic sub-demonstration optimality estimator are designed to promote disentanglement in the skill space. Extensive experiments are conducted over two gym environments and a real-world healthcare dataset to demonstrate the superiority of {\method} in learning from sub-optimal demonstrations and its improved interpretability by examining learned skills.