As automated decision making and decision assistance systems become common in everyday life, research on the prevention or mitigation of potential harms that arise from decisions made by these systems has proliferated. However, various research communities have independently conceptualized these harms, envisioned potential applications, and proposed interventions. The result is a somewhat fractured landscape of literature focused generally on ensuring decision-making algorithms "do the right thing". In this paper, we compare and discuss work across two major subsets of this literature: algorithmic fairness, which focuses primarily on predictive systems, and ethical decision making, which focuses primarily on sequential decision making and planning. We explore how each of these settings has articulated its normative concerns, the viability of different techniques for these different settings, and how ideas from each setting may have utility for the other.

Allowing agents to share information through communication is crucial for solving complex tasks in multi-agent reinforcement learning. In this work, we consider the question of whether a given communication protocol can express an arbitrary policy. By observing that many existing protocols can be viewed as instances of graph neural networks (GNNs), we demonstrate the equivalence of joint action selection to node labelling. With standard GNN approaches provably limited in their expressive capacity, we draw from existing GNN literature and consider augmenting agent observations with: (1) unique agent IDs and (2) random noise. We provide a theoretical analysis as to how these approaches yield universally expressive communication, and also prove them capable of targeting arbitrary sets of actions for identical agents. Empirically, these augmentations are found to improve performance on tasks where expressive communication is required, whilst, in general, the optimal communication protocol is found to be task-dependent.

This article explores the concepts of online protocol synthesis using Reinforcement Learning (RL). The study is performed in the context of sensor and IoT networks with ultra low complexity wireless transceivers. The paper introduces the use of RL and Multi Armed Bandit (MAB), a specific type of RL, for Medium Access Control (MAC) under different network and traffic conditions. It then introduces a novel learning based protocol synthesis framework that addresses specific difficulties and limitations in medium access for both random access and time slotted networks. The mechanism does not rely on carrier sensing, network time-synchronization, collision detection, and other low level complex operations, thus making it ideal for ultra simple transceiver hardware used in resource constrained sensor and IoT networks. Additionally, the ability of independent protocol learning by the nodes makes the system robust and adaptive to the changes in network and traffic conditions. It is shown that the nodes can be trained to learn to avoid collisions, and to achieve network throughputs that are comparable to ALOHA based access protocols in sensor and IoT networks with simplest transceiver hardware. It is also shown that using RL, it is feasible to synthesize access protocols that can sustain network throughput at high traffic loads, which is not feasible in the ALOHA-based systems. The ability of the system to provide throughput fairness under network and traffic heterogeneities are also experimentally demonstrated.

Testing Deep Learning (DL) systems is a complex task as they do not behave like traditional systems would, notably because of their stochastic nature. Nonetheless, being able to adapt existing testing techniques such as Mutation Testing (MT) to DL settings would greatly improve their potential verifiability. While some efforts have been made to extend MT to the Supervised Learning paradigm, little work has gone into extending it to Reinforcement Learning (RL) which is also an important component of the DL ecosystem but behaves very differently from SL. This paper builds on the existing approach of MT in order to propose a framework, RLMutation, for MT applied to RL. Notably, we use existing taxonomies of faults to build a set of mutation operators relevant to RL and use a simple heuristic to generate test cases for RL. This allows us to compare different mutation killing definitions based on existing approaches, as well as to analyze the behavior of the obtained mutation operators and their potential combinations called Higher Order Mutation(s) (HOM). We show that the design choice of the mutation killing definition can affect whether or not a mutation is killed as well as the generated test cases. Moreover, we found that even with a relatively small number of test cases and operators we manage to generate HOM with interesting properties which can enhance testing capability in RL systems.

Very large state spaces with a sparse reward signal are difficult to explore. The lack of a sophisticated guidance results in a poor performance for numerous reinforcement learning algorithms. In these cases, the commonly used random exploration is often not helpful. The literature shows that this kind of environments require enormous efforts to systematically explore large chunks of the state space. Learned state representations can help here to improve the search by providing semantic context and build a structure on top of the raw observations. In this work we introduce a novel time-myopic state representation that clusters temporal close states together while providing a time prediction capability between them. By adapting this model to the Go-Explore paradigm (Ecoffet et al., 2021b), we demonstrate the first learned state representation that reliably estimates novelty instead of using the hand-crafted representation heuristic. Our method shows an improved solution for the detachment problem which still remains an issue at the Go-Explore Exploration Phase. We provide evidence that our proposed method covers the entire state space with respect to all possible time trajectories without causing disadvantageous conflict-overlaps in the cell archive. Analogous to native Go-Explore, our approach is evaluated on the hard exploration environments MontezumaRevenge, Gravitar and Frostbite (Atari) in order to validate its capabilities on difficult tasks. Our experiments show that time-myopic Go-Explore is an effective alternative for the domain-engineered heuristic while also being more general. The source code of the method is available on GitHub.

We propose the first model-free algorithm that achieves low regret performance for decentralized learning in two-player zero-sum tabular stochastic games with infinite-horizon average-reward objective. In decentralized learning, the learning agent controls only one player and tries to achieve low regret performances against an arbitrary opponent. This contrasts with centralized learning where the agent tries to approximate the Nash equilibrium by controlling both players. In our infinite-horizon undiscounted setting, additional structure assumptions is needed to provide good behaviors of learning processes : here we assume for every strategy of the opponent, the agent has a way to go from any state to any other. This assumption is the analogous to the "communicating" assumption in the MDP setting. We show that our Decentralized Optimistic Nash Q-Learning (DONQ-learning) algorithm achieves both sublinear high probability regret of order $T^{3/4}$ and sublinear expected regret of order $T^{2/3}$. Moreover, our algorithm enjoys a low computational complexity and low memory space requirement compared to the previous works of (Wei et al. 2017) and (Jafarnia-Jahromi et al. 2021) in the same setting.

Medical images are generally acquired with limited field-of-view (FOV), which could lead to incomplete regions of interest (ROI), and thus impose a great challenge on medical image analysis. This is particularly evident for the learning-based multi-target landmark detection, where algorithms could be misleading to learn primarily the variation of background due to the varying FOV, failing the detection of targets. Based on learning a navigation policy, instead of predicting targets directly, reinforcement learning (RL)-based methods have the potential totackle this challenge in an efficient manner. Inspired by this, in this work we propose a multi-agent RL framework for simultaneous multi-target landmark detection. This framework is aimed to learn from incomplete or (and) complete images to form an implicit knowledge of global structure, which is consolidated during the training stage for the detection of targets from either complete or incomplete test images. To further explicitly exploit the global structural information from incomplete images, we propose to embed a shape model into the RL process. With this prior knowledge, the proposed RL model can not only localize dozens of targetssimultaneously, but also work effectively and robustly in the presence of incomplete images. We validated the applicability and efficacy of the proposed method on various multi-target detection tasks with incomplete images from practical clinics, using body dual-energy X-ray absorptiometry (DXA), cardiac MRI and head CT datasets. Results showed that our method could predict whole set of landmarks with incomplete training images up to 80% missing proportion (average distance error 2.29 cm on body DXA), and could detect unseen landmarks in regions with missing image information outside FOV of target images (average distance error 6.84 mm on 3D half-head CT).

5G New Radio proposes the usage of frequencies above 10 GHz to speed up LTE's existent maximum data rates. However, the effective size of 5G antennas and consequently its repercussions in the signal degradation in urban scenarios makes it a challenge to maintain stable coverage and connectivity. In order to obtain the best from both technologies, recent dual connectivity solutions have proved their capabilities to improve performance when compared with coexistent standalone 5G and 4G technologies. Reinforcement learning (RL) has shown its huge potential in wireless scenarios where parameter learning is required given the dynamic nature of such context. In this paper, we propose two reinforcement learning algorithms: a single agent RL algorithm named Clipped Double Q-Learning (CDQL) and a hierarchical Deep Q-Learning (HiDQL) to improve Multiple Radio Access Technology (multi-RAT) dual-connectivity handover. We compare our proposal with two baselines: a fixed parameter and a dynamic parameter solution. Simulation results reveal significant improvements in terms of latency with a gain of 47.6% and 26.1% for Digital-Analog beamforming (BF), 17.1% and 21.6% for Hybrid-Analog BF, and 24.7% and 39% for Analog-Analog BF when comparing the RL-schemes HiDQL and CDQL with the with the existent solutions, HiDQL presented a slower convergence time, however obtained a more optimal solution than CDQL. Additionally, we foresee the advantages of utilizing context-information as geo-location of the UEs to reduce the beam exploration sector, and thus improving further multi-RAT handover latency results.

In recent years, population ageing and pandemics have been demonstrated to cause isolation of older adults in their houses, generating the need for a reliable assistive figure. Service robotics recently emerged as high-tech support to the problem, providing a series of aid functionality to satisfy daily indoor assistance. Robotic solutions take care of interactive social aspects [1] or monitoring the health status of the user [2, 3]. Domestic environments are often very demanding for autonomous navigation systems due to the variety of complex and dynamic obstacles they can feature. To this end, the robot platform shall provide extreme flexibility and effective mobility to handle narrow passages thought for humans. Moreover, in order to properly assist the user, the platform should be able to follow them within this environment. Person following [4, 5] is the first step to enable any visual or vocal interaction with the user while monitoring its condition to intervene earlier in the case of anomalous events. Person following systems are often based on naive visual-control strategy, directly coupling the generation of heuristic commands for the robot with the person coordinate in the image [6]. Deep Reinforcement Learning (DRL) agents recently demonstrated significant autonomy and flexibility boost in robotic solutions.

We investigate whether self-supervised learning (SSL) can improve online reinforcement learning (RL) from pixels. We extend the contrastive reinforcement learning framework (e.g., CURL) that jointly optimizes SSL and RL losses and conduct an extensive amount of experiments with various self-supervised losses. Our observations suggest that the existing SSL framework for RL fails to bring meaningful improvement over the baselines only taking advantage of image augmentation when the same amount of data and augmentation is used. We further perform evolutionary searches to find the optimal combination of multiple self-supervised losses for RL, but find that even such a loss combination fails to meaningfully outperform the methods that only utilize carefully designed image augmentations. After evaluating these approaches together in multiple different environments including a real-world robot environment, we confirm that no single self-supervised loss or image augmentation method can dominate all environments and that the current framework for joint optimization of SSL and RL is limited. Finally, we conduct the ablation study on multiple factors and demonstrate the properties of representations learned with different approaches.