Learning policies which are robust to changes in the environment are critical for real world deployment of Reinforcement Learning agents. They are also necessary for achieving good generalization across environment shifts. We focus on bisimulation metrics, which provide a powerful means for abstracting task relevant components of the observation and learning a succinct representation space for training the agent using reinforcement learning. In this work, we extend the bisimulation framework to also account for context dependent observation shifts. Specifically, we focus on the simulator based learning setting and use alternate observations to learn a representation space which is invariant to observation shifts using a novel bisimulation based objective. This allows us to deploy the agent to varying observation settings during test time and generalize to unseen scenarios. We further provide novel theoretical bounds for simulator fidelity and performance transfer guarantees for using a learnt policy to unseen shifts. Empirical analysis on the high-dimensional image based control domains demonstrates the efficacy of our method.

The reinforcement learning algorithm SARSA combined with linear function approximation has been shown to converge for infinite horizon discounted Markov decision problems (MDPs). In this paper, we investigate the convergence of the algorithm for random horizon MDPs, which has not previously been shown. We show, similar to earlier results for infinite horizon discounted MDPs, that if the behaviour policy is $\varepsilon$-soft and Lipschitz continuous with respect to the weight vector of the linear function approximation, with small enough Lipschitz constant, then the algorithm will converge with probability one when considering a random horizon MDP.

The balancing process for game levels in a competitive two-player context involves a lot of manual work and testing, particularly in non-symmetrical game levels. In this paper, we propose an architecture for automated balancing of tile-based levels within the recently introduced PCGRL framework (procedural content generation via reinforcement learning). Our architecture is divided into three parts: (1) a level generator, (2) a balancing agent and, (3) a reward modeling simulation. By playing the level in a simulation repeatedly, the balancing agent is rewarded for modifying it towards the same win rates for all players. To this end, we introduce a novel family of swap-based representations to increase robustness towards playability. We show that this approach is capable to teach an agent how to alter a level for balancing better and faster than plain PCGRL. In addition, by analyzing the agent's swapping behavior, we can draw conclusions about which tile types influence the balancing most. We test and show our results using the Neural MMO (NMMO) environment in a competitive two-player setting.

The shift from the understanding and prediction of processes to their optimization offers great benefits to businesses and other organizations. Precisely timed process interventions are the cornerstones of effective optimization. Prescriptive process monitoring (PresPM) is the sub-field of process mining that concentrates on process optimization. The emerging PresPM literature identifies state-of-the-art methods, causal inference (CI) and reinforcement learning (RL), without presenting a quantitative comparison. Most experiments are carried out using historical data, causing problems with the accuracy of the methods' evaluations and preempting online RL. Our contribution consists of experiments on timed process interventions with synthetic data that renders genuine online RL and the comparison to CI possible, and allows for an accurate evaluation of the results. Our experiments reveal that RL's policies outperform those from CI and are more robust at the same time. Indeed, the RL policies approach perfect policies. Unlike CI, the unaltered online RL approach can be applied to other, more generic PresPM problems such as next best activity recommendations. Nonetheless, CI has its merits in settings where online learning is not an option.

Robots assist humans in various activities, from daily living public service (e.g., airports and restaurants), and to collaborative manufacturing. However, it is risky to assume that the knowledge and strategies robots learned from one group of people can apply to other groups. The discriminatory performance of robots will undermine their service quality for some people, ignore their service requests, and even offend them. Therefore, it is critically important to mitigate bias in robot decision-making for more fair services. In this paper, we designed a self-reflective mechanism -- Fairness-Sensitive Policy Gradient Reinforcement Learning (FSPGRL), to help robots to self-identify biased behaviors during interactions with humans. FSPGRL identifies bias by examining the abnormal update along particular gradients and updates the policy network to support fair decision-making for robots. To validate FSPGRL's effectiveness, a human-centered service scenario, "A robot is serving people in a restaurant," was designed. A user study was conducted; 24 human subjects participated in generating 1,000 service demonstrations. Four commonly-seen issues "Willingness Issue," "Priority Issue," "Quality Issue," "Risk Issue" were observed from robot behaviors. By using FSPGRL to improve robot decisions, robots were proven to have a self-bias detection capability for a more fair service. We have achieved the suppression of bias and improved the quality during the process of robot learning to realize a relatively fair model.

The development of urban-air-mobility (UAM) is rapidly progressing with spurs, and the demand for efficient transportation management systems is a rising need due to the multifaceted environmental uncertainties. Thus, this paper proposes a novel air transportation service management algorithm based on multi-agent deep reinforcement learning (MADRL) to address the challenges of multi-UAM cooperation. Specifically, the proposed algorithm in this paper is based on communication network (CommNet) method utilizing centralized training and distributed execution (CTDE) in multiple UAMs for providing efficient air transportation services to passengers collaboratively. Furthermore, this paper adopts actual vertiport maps and UAM specifications for constructing realistic air transportation networks. By evaluating the performance of the proposed algorithm in data-intensive simulations, the results show that the proposed algorithm outperforms existing approaches in terms of air transportation service quality. Furthermore, there are no inferior UAMs by utilizing parameter sharing in CommNet and a centralized critic network in CTDE. Therefore, it can be confirmed that the research results in this paper can provide a promising solution for autonomous air transportation management systems in city-wide urban areas.

Pretraining RL models on offline video datasets is a promising way to improve their training efficiency in online tasks, but challenging due to the inherent mismatch in tasks, dynamics, and behaviors across domains. A recent model, APV [48], sidesteps the accompanied action records in offline datasets and instead focuses on pretraining a task-irrelevant, action-free world model within the source domains. We present Vid2Act, a model-based RL method that learns to transfer valuable action-conditioned dynamics and potentially useful action demonstrations from offline to online settings. The main idea is to use the world models not only as simulators for behavior learning but also as tools to measure the domain relevance for both dynamics representation transfer and policy transfer. Specifically, we train the world models to generate a set of time-varying task similarities using a domain-selective knowledge distillation loss. These similarities serve two purposes: (i) adaptively transferring the most useful source knowledge to facilitate dynamics learning, and (ii) learning to replay the most relevant source actions to guide the target policy. We demonstrate the advantages of Vid2Act over the action-free visual RL pretraining method [48] in both Meta-World and DeepMind Control Suite.

A number of deep reinforcement-learning (RL) approaches propose to control traffic signals. In this work, we study the robustness of such methods along two axes. First, sensor failures and GPS occlusions create missing-data challenges and we show that recent methods remain brittle in the face of these missing data. Second, we provide a more systematic study of the generalization ability of RL methods to new networks with different traffic regimes. Again, we identify the limitations of recent approaches. We then propose using a combination of distributional and vanilla reinforcement learning through a policy ensemble. Building upon the state-of-the-art previous model which uses a decentralized approach for large-scale traffic signal control with graph convolutional networks (GCNs), we first learn models using a distributional reinforcement learning (DisRL) approach. In particular, we use implicit quantile networks (IQN) to model the state-action return distribution with quantile regression. For traffic signal control problems, an ensemble of standard RL and DisRL yields superior performance across different scenarios, including different levels of missing sensor data and traffic flow patterns. Furthermore, the learning scheme of the resulting model can improve zero-shot transferability to different road network structures, including both synthetic networks and real-world networks (e.g., Luxembourg, Manhattan). We conduct extensive experiments to compare our approach to multi-agent reinforcement learning and traditional transportation approaches. Results show that the proposed method improves robustness and generalizability in the face of missing data, varying road networks, and traffic flows.

Value decomposition is widely used in cooperative multi-agent reinforcement learning, however, its implicit credit assignment mechanism is not yet fully understood due to black-box networks. In this work, we study an interpretable value decomposition framework via the family of generalized additive models. We present a novel method, named Neural Attention Additive Q-learning (N$\text{A}^\text{2}$Q), providing inherent intelligibility of collaboration behavior. N$\text{A}^\text{2}$Q can explicitly factorize the optimal joint policy induced by enriching shape functions to model all possible coalitions of agents into individual policies. Moreover, we construct identity semantics to promote estimating credits together with the global state and individual value functions, where local semantic masks help us diagnose whether each agent captures relevant-task information. Extensive experiments show that N$\text{A}^\text{2}$Q consistently achieves superior performance compared to different state-of-the-art methods on all challenging tasks, while yielding human-like interpretability.

As machine learning (ML) algorithms are deployed for tasks with real-world social consequences (e.g., school admissions, loan approval, medical interventions, etc.), the possibility exists for runaway social inequalities (Crawford and Calo, 2016; Chaney et al., 2018; Fuster et al., 2018; Ensign et al., 2018). While "fairness" has become a salient ethical concern in contemporary research, the closed-loop dynamics of real-world systems comprising ML policies and populations that mutually adapt to each other (Figure 1 in the supplementary material) remain poorly understood. In this paper, our primary contribution is to consider the problem of long-term fairness, or algorithmic fairness in the context of a dynamically responsive population, as a reinforcement learning (RL) problem subject to constraint. In our formulation, the central learning task is to develop a policy that minimizes cumulative loss (e.g., financial risk, negative educational outcomes, misdiagnoses, etc.) incurred by an ML agent interacting with a human population up to a finite time horizon, subject to constraints on cumulative "violations of fairness", which we refer to in a single time step as disparity and cumulatively as distortion.