Decentralized cooperation in partially-observable multi-agent systems requires effective communications among agents. To support this effort, this work focuses on the class of problems where global communications are available but may be unreliable, thus precluding differentiable communication learning methods. We introduce FCMNet, a reinforcement learning based approach that allows agents to simultaneously learn a) an effective multi-hop communications protocol and b) a common, decentralized policy that enables team-level decision-making. Specifically, our proposed method utilizes the hidden states of multiple directional recurrent neural networks as communication messages among agents. Using a simple multi-hop topology, we endow each agent with the ability to receive information sequentially encoded by every other agent at each time step, leading to improved global cooperation. We demonstrate FCMNet on a challenging set of StarCraft II micromanagement tasks with shared rewards, as well as a collaborative multi-agent pathfinding task with individual rewards. There, our comparison results show that FCMNet outperforms state-of-the-art communication-based reinforcement learning methods in all StarCraft II micromanagement tasks, and value decomposition methods in certain tasks. We further investigate the robustness of FCMNet under realistic communication disturbances, such as random message loss or binarized messages (i.e., non-differentiable communication channels), to showcase FMCNet's potential applicability to robotic tasks under a variety of real-world conditions.
Serially connected robots are promising candidates for performing tasks in confined spaces such as search-and-rescue in large-scale disasters. Such robots are typically limbless, and we hypothesize that the addition of limbs could improve mobility. However, a challenge in designing and controlling such devices lies in the coordination of high-dimensional redundant modules in a way that improves mobility. Here we develop a general framework to control serially connected multi-legged robots. Specifically, we combine two approaches to build a general shape control scheme which can provide baseline patterns of self-deformation ("gaits") for effective locomotion in diverse robot morphologies. First, we take inspiration from a dimensionality reduction and a biological gait classification scheme to generate cyclic patterns of body deformation and foot lifting/lowering, which facilitate generation of arbitrary substrate contact patterns. Second, we use geometric mechanics methods to facilitates identification of optimal phasing of these undulations to maximize speed and/or stability. Our scheme allows the development of effective gaits in multi-legged robots locomoting on flat frictional terrain with diverse number of limbs (4, 6, 16, and even 0 limbs) and body actuation capabilities (including sidewinding gaits on limbless devices). By properly coordinating the body undulation and the leg placement, our framework combines the advantages of both limbless robots (modularity) and legged robots (mobility). We expect that our framework can provide general control schemes for the rapid deployment of general multi-legged robots, paving the ways toward machines that can traverse complex environments under real-life conditions.
The multiple traveling salesman problem (mTSP) is a well-known NP-hard problem with numerous real-world applications. In particular, this work addresses MinMax mTSP, where the objective is to minimize the max tour length (sum of Euclidean distances) among all agents. The mTSP is normally considered as a combinatorial optimization problem, but due to its computational complexity, search-based exact and heuristic algorithms become inefficient as the number of cities increases. Encouraged by the recent developments in deep reinforcement learning (dRL), this work considers the mTSP as a cooperative task and introduces a decentralized attention-based neural network method to solve the MinMax mTSP, named DAN. In DAN, agents learn fully decentralized policies to collaboratively construct a tour, by predicting the future decisions of other agents. Our model relies on the Transformer architecture, and is trained using multi-agent RL with parameter sharing, which provides natural scalability to the numbers of agents and cities. We experimentally demonstrate our model on small- to large-scale mTSP instances, which involve 50 to 1000 cities and 5 to 20 agents, and compare against state-of-the-art baselines. For small-scale problems (fewer than 100 cities), DAN is able to closely match the performance of the best solver available (OR Tools, a meta-heuristic solver) given the same computation time budget. In larger-scale instances, DAN outperforms both conventional and dRL-based solvers, while keeping computation times low, and exhibits enhanced collaboration among agents.
The Flatland competition aimed at finding novel approaches to solve the vehicle re-scheduling problem (VRSP). The VRSP is concerned with scheduling trips in traffic networks and the re-scheduling of vehicles when disruptions occur, for example the breakdown of a vehicle. While solving the VRSP in various settings has been an active area in operations research (OR) for decades, the ever-growing complexity of modern railway networks makes dynamic real-time scheduling of traffic virtually impossible. Recently, multi-agent reinforcement learning (MARL) has successfully tackled challenging tasks where many agents need to be coordinated, such as multiplayer video games. However, the coordination of hundreds of agents in a real-life setting like a railway network remains challenging and the Flatland environment used for the competition models these real-world properties in a simplified manner. Submissions had to bring as many trains (agents) to their target stations in as little time as possible. While the best submissions were in the OR category, participants found many promising MARL approaches. Using both centralized and decentralized learning based approaches, top submissions used graph representations of the environment to construct tree-based observations. Further, different coordination mechanisms were implemented, such as communication and prioritization between agents. This paper presents the competition setup, four outstanding solutions to the competition, and a cross-comparison between them.
Efficient automated scheduling of trains remains a major challenge for modern railway systems. The underlying vehicle rescheduling problem (VRSP) has been a major focus of Operations Research (OR) since decades. Traditional approaches use complex simulators to study VRSP, where experimenting with a broad range of novel ideas is time consuming and has a huge computational overhead. In this paper, we introduce a two-dimensional simplified grid environment called "Flatland" that allows for faster experimentation. Flatland does not only reduce the complexity of the full physical simulation, but also provides an easy-to-use interface to test novel approaches for the VRSP, such as Reinforcement Learning (RL) and Imitation Learning (IL). In order to probe the potential of Machine Learning (ML) research on Flatland, we (1) ran a first series of RL and IL experiments and (2) design and executed a public Benchmark at NeurIPS 2020 to engage a large community of researchers to work on this problem. Our own experimental results, on the one hand, demonstrate that ML has potential in solving the VRSP on Flatland. On the other hand, we identify key topics that need further research. Overall, the Flatland environment has proven to be a robust and valuable framework to investigate the VRSP for railway networks. Our experiments provide a good starting point for further research and for the participants of the NeurIPS 2020 Flatland Benchmark. All of these efforts together have the potential to have a substantial impact on shaping the mobility of the future.
Multi-agent path finding (MAPF) is an indispensable component of large-scale robot deployments in numerous domains ranging from airport management to warehouse automation. In particular, this work addresses lifelong MAPF (LMAPF) -- an online variant of the problem where agents are immediately assigned a new goal upon reaching their current one -- in dense and highly structured environments, typical of real-world warehouses operations. Effectively solving LMAPF in such environments requires expensive coordination between agents as well as frequent replanning abilities, a daunting task for existing coupled and decoupled approaches alike. With the purpose of achieving considerable agent coordination without any compromise on reactivity and scalability, we introduce PRIMAL2, a distributed reinforcement learning framework for LMAPF where agents learn fully decentralized policies to reactively plan paths online in a partially observable world. We extend our previous work, which was effective in low-density sparsely occupied worlds, to highly structured and constrained worlds by identifying behaviors and conventions which improve implicit agent coordination, and enabling their learning through the construction of a novel local agent observation and various training aids. We present extensive results of PRIMAL2 in both MAPF and LMAPF environments with up to 1024 agents and compare its performance to complete state-of-the-art planners. We experimentally observe that agents successfully learn to follow ideal conventions and can exhibit selfless coordinated maneuvers that maximize joint rewards. We find that not only does PRIMAL2 significantly surpass our previous work, it is also able to perform on par and even outperform state-of-the-art planners in terms of throughput.
Multi-agent foraging (MAF) involves distributing a team of agents to search an environment and extract resources from it. Many foraging algorithms use biologically-inspired signaling mechanisms, such as pheromones, to help agents navigate from resources back to a central nest while relying on local sensing only. However, these approaches often rely on predictable pheromone dynamics and/or perfect robot localization. In nature, certain environmental factors (e.g., heat or rain) can disturb or destroy pheromone trails, while imperfect sensing can lead robots astray. In this work, we propose ForMIC, a distributed reinforcement learning MAF approach that relies on pheromones as a way to endow agents with implicit communication abilities via their shared environment. Specifically, full agents involuntarily lay trails of pheromones as they move; other agents can then measure the local levels of pheromones to guide their individual decisions. We show how these stigmergic interactions among agents can lead to a highly-scalable, decentralized MAF policy that is naturally resilient to common environmental disturbances, such as depleting resources and sudden pheromone disappearance. We present simulation results that compare our learning policy against existing state-of-the-art MAF algorithms, in a set of experiments varying team sizes, number and placement of resources, and key environmental disturbances. Our results demonstrate that our learned policy outperforms these baselines, approaching the performance of a planner with full observability and centralized agent allocation. Preprint of the paper submitted to the IEEE Transactions on Robotics (T-RO) journal's special issue on Resilience in Networked Robotic Systems in June 2020
Multi-agent path finding (MAPF) is an essential component of many large-scale, real-world robot deployments, from aerial swarms to warehouse automation. However, despite the community's continued efforts, most state-of-the-art MAPF planners still rely on centralized planning and scale poorly past a few hundred agents. Such planning approaches are maladapted to real-world deployments, where noise and uncertainty often require paths be recomputed online, which is impossible when planning times are in seconds to minutes. We present PRIMAL, a novel framework for MAPF that combines reinforcement and imitation learning to teach fully-decentralized policies, where agents reactively plan paths online in a partially-observable world while exhibiting implicit coordination. This framework extends our previous work on distributed learning of collaborative policies by introducing demonstrations of an expert MAPF planner during training, as well as careful reward shaping and environment sampling. Once learned, the resulting policy can be copied onto any number of agents and naturally scales to different team sizes and world dimensions. We present results on randomized worlds with up to 1024 agents and compare success rates against state-of-the-art MAPF planners. Finally, we experimentally validate the learned policies in a hybrid simulation of a factory mockup, involving both real-world and simulated robots.
State-of-the-art distributed algorithms for reinforcement learning rely on multiple independent agents, which simultaneously learn in parallel environments while asynchronously updating a common, shared policy. Moreover, decentralized control architectures (e.g., CPGs) can coordinate spatially distributed portions of an articulated robot to achieve system-level objectives. In this work, we investigate the relationship between distributed learning and decentralized control by learning decentralized control policies for the locomotion of articulated robots in challenging environments. To this end, we present an approach that leverages the structure of the asynchronous advantage actor-critic (A3C) algorithm to provide a natural means of learning decentralized control policies on a single articulated robot. Our primary contribution shows individual agents in the A3C algorithm can be defined by independently controlled portions of the robot's body, thus enabling distributed learning on a single robot for efficient hardware implementation. We present results of closed-loop locomotion in unstructured terrains on a snake and a hexapod robot, using decentralized controllers learned offline and online respectively. Preprint of the paper submitted to the IEEE Transactions in Robotics (T-RO) journal in October 2018, and conditionally accepted for publication as a regular paper in January 2019.
Recent literature in the robotics community has focused on learning robot behaviors that abstract out lower-level details of robot control. To fully leverage the efficacy of such behaviors, it is necessary to select and sequence them to achieve a given task. In this paper, we present an approach to both learn and sequence robot behaviors, applied to the problem of visual navigation of mobile robots. We construct a layered representation of control policies composed of low- level behaviors and a meta-level policy. The low-level behaviors enable the robot to locomote in a particular environment while avoiding obstacles, and the meta-level policy actively selects the low-level behavior most appropriate for the current situation based purely on visual feedback. We demonstrate the effectiveness of our method on three simulated robot navigation tasks: a legged hexapod robot which must successfully traverse varying terrain, a wheeled robot which must navigate a maze-like course while avoiding obstacles, and finally a wheeled robot navigating in the presence of dynamic obstacles. We show that by learning control policies in a layered manner, we gain the ability to successfully traverse new compound environments composed of distinct sub-environments, and outperform both the low-level behaviors in their respective sub-environments, as well as a hand-crafted selection of low-level policies on these compound environments.