Reinforcement learning (RL), while being the benchmark for policy formulation, often struggles to deliver robust solutions across varying scenarios, leading to marked performance drops under environmental perturbations. Traditional adversarial training, based on a two-player max-min game, is known to bolster the robustness of RL agents, but it faces challenges: first, the complexity of the worst-case optimization problem may induce \emph{over-optimism}, and second, the choice of a specific set of potential adversaries might lead to \emph{over-pessimism} by considering implausible scenarios. In this work, we first observe that these two challenges do not balance out each other. Thus, we propose to apply variational optimization to optimize over the worst-case distribution of the adversary instead of a single worst-case adversary. Moreover, to counteract over-optimism, we train the RL agent to maximize the lower quantile of the cumulative rewards under worst-case adversary distribution. Our novel algorithm demonstrates a significant advancement over existing robust RL methods, corroborating the importance of the identified challenges and the effectiveness of our approach. To alleviate computational overhead associated with the proposed approach, we also propose a simplified version with lower computational burden and only minimal performance degradation. Extensive experiments validate that our approaches consistently yield policies with superior robustness.

In the realm of reinforcement learning (RL), off-policy evaluation (OPE) holds a pivotal position, especially in high-stake human-involved scenarios such as e-learning and healthcare. Applying OPE to these domains is often challenging with scarce and underrepresentative offline training trajectories. Data augmentation has been a successful technique to enrich training data. However, directly employing existing data augmentation methods to OPE may not be feasible, due to the Markovian nature within the offline trajectories and the desire for generalizability across diverse target policies. In this work, we propose an offline trajectory augmentation approach to specifically facilitate OPE in human-involved scenarios. We propose sub-trajectory mining to extract potentially valuable sub-trajectories from offline data, and diversify the behaviors within those sub-trajectories by varying coverage of the state-action space. Our work was empirically evaluated in a wide array of environments, encompassing both simulated scenarios and real-world domains like robotic control, healthcare, and e-learning, where the training trajectories include varying levels of coverage of the state-action space. By enhancing the performance of a variety of OPE methods, our work offers a promising path forward for tackling OPE challenges in situations where data may be limited or underrepresentative.

Decision Transformers (DTs) have been highly effective for offline reinforcement learning (RL) tasks, successfully modeling the sequences of actions in a given set of demonstrations. However, DTs may perform poorly in stochastic environments, which are prevalent in robotics scenarios. In this paper, we identify that the root cause of this performance degradation is the growing variance of returns-to-go, the signal used by DTs to predict actions, accumulated over the horizon. Building upon this insight, we propose an extension to DTs that allows them to be steered toward high-reward regions, where the expected returns are estimated using temporal difference learning. This way, we not only mitigate the growing variance problem but also eliminate the need for DTs to have access to returns-to-go during evaluation and deployment phases. We show that our method outperforms state-of-the-art offline RL methods in both simulated and real-world robotic arm environments.

In human-centric applications like healthcare and education, the \textit{heterogeneity} among patients and students necessitates personalized treatments and instructional interventions. While reinforcement learning (RL) has been utilized in those tasks, off-policy selection (OPS) is pivotal to close the loop by offline evaluating and selecting policies without online interactions, yet current OPS methods often overlook the heterogeneity among participants. Our work is centered on resolving a \textit{pivotal challenge} in human-centric systems (HCSs): \textbf{\textit{how to select a policy to deploy when a new participant joining the cohort, without having access to any prior offline data collected over the participant?}} We introduce First-Glance Off-Policy Selection (FPS), a novel approach that systematically addresses participant heterogeneity through sub-group segmentation and tailored OPS criteria to each sub-group. By grouping individuals with similar traits, FPS facilitates personalized policy selection aligned with unique characteristics of each participant or group of participants. FPS is evaluated via two important but challenging applications, intelligent tutoring systems and a healthcare application for sepsis treatment and intervention. FPS presents significant advancement in enhancing learning outcomes of students and in-hospital care outcomes.

Off-policy evaluation (OPE) is important for closing the gap between offline training and evaluation of reinforcement learning (RL), by estimating performance and/or rank of target (evaluation) policies using offline trajectories only. It can improve the safety and efficiency of data collection and policy testing procedures in situations where online deployments are expensive, such as healthcare. However, existing OPE methods fall short in estimating human feedback (HF) signals, as HF may be conditioned over multiple underlying factors and are only sparsely available; as opposed to the agent-defined environmental rewards (used in policy optimization), which are usually determined over parametric functions or distributions. Consequently, the nature of HF signals makes extrapolating accurate OPE estimations to be challenging. To resolve this, we introduce an OPE for HF (OPEHF) framework that revives existing OPE methods in order to accurately evaluate the HF signals. Specifically, we develop an immediate human reward (IHR) reconstruction approach, regularized by environmental knowledge distilled in a latent space that captures the underlying dynamics of state transitions as well as issuing HF signals. Our approach has been tested over *two real-world experiments*, adaptive *in-vivo* neurostimulation and intelligent tutoring, and a simulation environment (visual Q&A). Results show that our approach significantly improves the performance toward estimating HF signals accurately, compared to directly applying (variants of) existing OPE methods.

Multivariate time series (MTS) classification is a challenging and important task in various domains and real-world applications. Much of prior work on MTS can be roughly divided into neural network (NN)- and pattern-based methods. The former can lead to robust classification performance, but many of the generated patterns are challenging to interpret; while the latter often produce interpretable patterns that may not be helpful for the classification task. In this work, we propose a reinforcement learning (RL) informed PAttern Mining framework (RLPAM) to identify interpretable yet important patterns for MTS classification. Our framework has been validated by 30 benchmark datasets as well as real-world large-scale electronic health records (EHRs) for an extremely challenging task: sepsis shock early prediction. We show that RLPAM outperforms the state-of-the-art NN-based methods on 14 out of 30 datasets as well as on the EHRs. Finally, we show how RL informed patterns can be interpretable and can improve our understanding of septic shock progression.

In this paper, we propose a Deep Imitative Q-learning (DIQL) method to synthesize control policies for mobile robots that need to satisfy Linear Temporal Logic (LTL) specifications using noisy semantic observations of their surroundings. The robot sensing error is modeled using probabilistic labels defined over the states of a Labeled Transition System (LTS) and the robot mobility is modeled using a Labeled Markov Decision Process (LMDP) with unknown transition probabilities. We use existing product-based model checkers (PMCs) as experts to guide the Q-learning algorithm to convergence. To the best of our knowledge, this is the first approach that models noise in semantic observations using probabilistic labeling functions and employs existing model checkers to provide suboptimal instructions to the Q-learning agent.