It is well-known that ensemble methods often provide enhanced performance in reinforcement learning. In this paper we explore this concept further by using group-aided training within the distributional reinforcement learning paradigm. Specifically, we propose an extension to categorical reinforcement learning, where distributional learning targets are implicitly based on the total information gathered by an ensemble. We empirically show that this may lead to much more robust initial learning, a stronger individual performance level and good efficiency on a per-sample basis.

Recent advances in machine learning are consistently enabled by increasing amounts of computation. Reinforcement learning (RL) and population-based methods in particular pose unique challenges for efficiency and flexibility to the underlying distributed computing frameworks. These challenges include frequent interaction with simulations, the need for dynamic scaling, and the need for a user interface with low adoption cost and consistency across different backends. In this paper we address these challenges while still retaining development efficiency and flexibility for both research and practical applications by introducing Fiber, a scalable distributed computing framework for RL and population-based methods. Fiber aims to significantly expand the accessibility of large-scale parallel computation to users of otherwise complicated RL and population-based approaches without the need to for specialized computational expertise.

Reinforcement learning (RL) has proven its worth in a series of artificial domains, and is beginning to show some successes in real-world scenarios. However, much of the research advances in RL are hard to leverage in real-world systems due to a series of assumptions that are rarely satisfied in practice. In this work, we identify and formalize a series of independent challenges that embody the difficulties that must be addressed for RL to be commonly deployed in real-world systems. For each challenge, we define it formally in the context of a Markov Decision Process, analyze the effects of the challenge on state-of-the-art learning algorithms, and present some existing attempts at tackling it. We believe that an approach that addresses our set of proposed challenges would be readily deployable in a large number of real world problems. Our proposed challenges are implemented in a suite of continuous control environments called realworldrl-suite which we propose an as an open-source benchmark.

Off-policy estimation for long-horizon problems is important in many real-life applications such as healthcare and robotics, where high-fidelity simulators may not be available and on-policy evaluation is expensive or impossible. Recently, \cite{liu18breaking} proposed an approach that avoids the \emph{curse of horizon} suffered by typical importance-sampling-based methods. While showing promising results, this approach is limited in practice as it requires data be drawn from the \emph{stationary distribution} of a \emph{known} behavior policy. In this work, we propose a novel approach that eliminates such limitations. In particular, we formulate the problem as solving for the fixed point of a certain operator. Using tools from Reproducing Kernel Hilbert Spaces (RKHSs), we develop a new estimator that computes importance ratios of stationary distributions, without knowledge of how the off-policy data are collected. We analyze its asymptotic consistency and finite-sample generalization. Experiments on benchmarks verify the effectiveness of our approach.

This work presents an application of Reinforcement Learning (RL) for the complete control of real soccer robots of the IEEE Very Small Size Soccer (VSSS) [1], a traditional league in the Latin American Robotics Competition (LARC). In the VSSS league, two teams of three small robots play against each other. We propose a simulated environment in which continuous or discrete control policies can be trained, and a Sim-to-Real method to allow using the obtained policies to control a robot in the real world. The results show that the learned policies display a broad repertoire of behaviors which are difficult to specify by hand. This approach, called VSSS-RL, was able to beat the human-designed policy for the striker of the team ranked 3rd place in the 2018 LARC, in 1-vs-1 matches.

The new system, dubbed DeepSPM, bridges the gap between nanoscience, automation and artificial intelligence (AI), and firmly establishes the use of machine learning for experimental scientific research. "Optimising SPM data acquisition can be very tedious. This optimisation process is usually performed by the human experimentalist, and is rarely reported," says FLEET Chief Investigator Dr Agustin Schiffrin (Monash University). "Our new AI-driven system can operate and acquire optimal SPM data autonomously, for multiple straight days, and without any human supervision." The advance brings advanced SPM methodologies such as atomically-precise nanofabrication and high-throughput data acquisition closer to a fully automated turnkey application.

Deep neural networks have been widely adopted in modern reinforcement learning (RL) algorithms with great empirical successes in various domains. However, the large search space of training a neural network requires a significant amount of data, which makes the current RL algorithms not sample efficient. Motivated by the fact that many environments with continuous state space have smooth transitions, we propose to learn a smooth policy that behaves smoothly with respect to states. In contrast to policies parameterized by linear/reproducing kernel functions, where simple regularization techniques suffice to control smoothness, for neural network based reinforcement learning algorithms, there is no readily available solution to learn a smooth policy. In this paper, we develop a new training framework --- $\textbf{S}$mooth $\textbf{R}$egularized $\textbf{R}$einforcement $\textbf{L}$earning ($\textbf{SR}^2\textbf{L}$), where the policy is trained with smoothness-inducing regularization. Such regularization effectively constrains the search space of the learning algorithms and enforces smoothness in the learned policy. We apply the proposed framework to both on-policy (TRPO) and off-policy algorithm (DDPG). Through extensive experiments, we demonstrate that our method achieves improved sample efficiency.

Many cooperative multi-agent problems require agents to learn individual tasks while contributing to the collective success of the group. This is a challenging task for current state-of-the-art multi-agent reinforcement algorithms that are designed to either maximize the global reward of the team or the individual local rewards. The problem is exacerbated when either of the rewards is sparse leading to unstable learning. To address this problem, we present Decomposed Multi-Agent Deep Deterministic Policy Gradient (DE-MADDPG): a novel cooperative multi-agent reinforcement learning framework that simultaneously learns to maximize the global and local rewards. We evaluate our solution on the challenging defensive escort team problem and show that our solution achieves a significantly better and more stable performance than the direct adaptation of the MADDPG algorithm.

In multi-agent games, the complexity of the environment can grow exponentially as the number of agents increases, so it is particularly challenging to learn good policies when the agent population is large. In this paper, we introduce Evolutionary Population Curriculum (EPC), a curriculum learning paradigm that scales up Multi-Agent Reinforcement Learning (MARL) by progressively increasing the population of training agents in a stage-wise manner. Furthermore, EPC uses an evolutionary approach to fix an objective misalignment issue throughout the curriculum: agents successfully trained in an early stage with a small population are not necessarily the best candidates for adapting to later stages with scaled populations. Concretely, EPC maintains multiple sets of agents in each stage, performs mix-and-match and fine-tuning over these sets and promotes the sets of agents with the best adaptability to the next stage. We implement EPC on a popular MARL algorithm, MADDPG, and empirically show that our approach consistently outperforms baselines by a large margin as the number of agents grows exponentially. The project page is https://sites.google.com/view/epciclr2020.

Unmanned Surface Vehicles technology (USVs) is an exciting topic that essentially deploys an algorithm to safely and efficiently performs a mission. Although reinforcement learning is a well-known approach to modeling such a task, instability and divergence may occur when combining off-policy and function approximation. In this work, we used deep reinforcement learning combining Q-learning with a neural representation to avoid instability. Our methodology uses deep q-learning and combines it with a rolling wave planning approach on agile methodology. Our method contains two critical parts in order to perform missions in an unknown environment. The first is a path planner that is responsible for generating a potential effective path to a destination without considering the details of the root. The latter is a decision-making module that is responsible for short-term decisions on avoiding obstacles during the near future steps of USV exploitation within the context of the value function. Simulations were performed using two algorithms: a basic vanilla vessel navigator (VVN) as a baseline and an improved one for the vessel navigator with a planner and local view (VNPLV). Experimental results show that the proposed method enhanced the performance of VVN by 55.31 on average for long-distance missions. Our model successfully demonstrated obstacle avoidance by means of deep reinforcement learning using planning adaptive paths in unknown environments.