Guillaume Sartoretti

Xinrong Yang, Peizhuo Li, Hongyi Li et al.

In nature, animals often need to move/manipulate objects comparable in weight/size to their own bodies. Compared to grasping and carrying, pushing provides a more straightforward and efficient non-prehensile manipulation strategy, avoiding complex grasp design while leveraging direct contact to regulate an object's pose during interaction. Achieving effective pushing, however, requires both sufficient manipulation capability and stable whole-body coordination, which is particularly challenging when dealing with heavy or irregular objects. To address these challenges, we propose HeLoM, a learning-based hierarchical whole-body manipulation framework for hexapod robots that exploits coordinated multi-limb control and is applicable to multi-legged robotic systems. Inspired by the cooperative strategies of multi-legged insects, our framework leverages multiple contact points and high degrees of freedom to enable efficient and dynamic whole-body coordination during object interaction. HeLoM's high-level planner plans pushing behaviors, while its low-level controller maintains locomotion stability and generates dynamically consistent joint actions. This design enables the robot to maintain balance while executing continuous and controllable pushing behaviors through coordinated foreleg interaction and supportive hind-leg propulsion. We validate the effectiveness of HeLoM through both simulation and real-world experiments. Results show that our framework can stably push objects of varying sizes and unknown physical properties to designated goal poses in the real world.

Jeric Lew, Yuhong Cao, Derek Ming Siang Tan et al.

Information gathering in large-scale or time-critical scenarios (e.g., environmental monitoring, search and rescue) requires broad coverage within limited time budgets, motivating the use of multi-agent systems. These scenarios are commonly formulated as multi-agent informative path planning (MAIPP), where multiple agents must coordinate to maximize information gain while operating under budget constraints. A central challenge in MAIPP is ensuring effective coordination while the belief over the environment evolves with incoming measurements. Recent learning-based approaches address this by using distributions over future positions as "intent" to support coordination. However, these autoregressive intent predictors are computationally expensive and prone to compounding errors. Inspired by the effectiveness of diffusion models as expressive, long-horizon policies, we propose AID, a fully decentralized MAIPP framework that leverages diffusion models to generate long-term trajectories in a non-autoregressive manner. AID first performs behavior cloning on trajectories produced by existing MAIPP planners and then fine-tunes the policy using reinforcement learning via Diffusion Policy Policy Optimization (DPPO). This two-stage pipeline enables the policy to inherit expert behavior while learning improved coordination through online reward feedback. Experiments demonstrate that AID consistently improves upon the MAIPP planners it is trained from, achieving 4x faster execution and up to 17% increased information gain, while scaling effectively to larger numbers of agents. Our implementation is publicly available at https://github.com/marmotlab/AID.

Yifeng Zhang, Harsh Goel, Peizhuo Li et al.

Adaptive traffic signal control (ATSC) is crucial in alleviating congestion, maximizing throughput and promoting sustainable mobility in ever-expanding cities. Multi-Agent Reinforcement Learning (MARL) has recently shown significant potential in addressing complex traffic dynamics, but the intricacies of partial observability and coordination in decentralized environments still remain key challenges in formulating scalable and efficient control strategies. To address these challenges, we present CoordLight, a MARL-based framework designed to improve intra-neighborhood traffic by enhancing decision-making at individual junctions (agents), as well as coordination with neighboring agents, thereby scaling up to network-level traffic optimization. Specifically, we introduce the Queue Dynamic State Encoding (QDSE), a novel state representation based on vehicle queuing models, which strengthens the agents' capability to analyze, predict, and respond to local traffic dynamics. We further propose an advanced MARL algorithm, named Neighbor-aware Policy Optimization (NAPO). It integrates an attention mechanism that discerns the state and action dependencies among adjacent agents, aiming to facilitate more coordinated decision-making, and to improve policy learning updates through robust advantage calculation. This enables agents to identify and prioritize crucial interactions with influential neighbors, thus enhancing the targeted coordination and collaboration among agents. Through comprehensive evaluations against state-of-the-art traffic signal control methods over three real-world traffic datasets composed of up to 196 intersections, we empirically show that CoordLight consistently exhibits superior performance across diverse traffic networks with varying traffic flows. The code is available at https://github.com/marmotlab/CoordLight

Shizhe Zhang, Jingsong Liang, Zhitao Zhou et al.

Existing methods for multi-agent navigation typically assume fully known environments, offering limited support for partially known scenarios with outdated or imperfect prior maps, such as warehouses or factory floors. There, agents need to balance path optimality with collecting and sharing environmental information to help teammates reach their own targets. To these ends, we propose ORION, a novel deep reinforcement learning framework for cooperative multi-agent online navigation in partially known environments. Starting from an imperfect prior map, ORION trains agents to make decentralized decisions, coordinate toward individual targets, and actively reduce task-relevant map uncertainty through online observation sharing in a closed perception-action loop. We first design a shared graph encoder that fuses prior map with online perception into a unified representation, providing robust state embeddings under environmental discrepancies. At the core of ORION is an option-critic framework that learns high-level cooperative modes translated into sequences of low-level actions, enabling adaptive switching between individual navigation and team-level exploration. We further introduce a dual-stage cooperation strategy that allows agents to assist teammates under map uncertainty, thereby reducing the overall makespan. Across extensive maze-like maps and large-scale warehouse environments, ORION achieves high-quality real-time decentralized cooperation while scaling to up to 10 robots, outperforming state-of-the-art classical and learning-based baselines. Finally, we validate ORION on physical robot teams, demonstrating its robustness and practicality for real-world cooperative navigation.

Yutong Wang, Mehul Damani, Pamela Wang et al.

Purpose of review: Recent advances in sensing, actuation, and computation have opened the door to multi-robot systems consisting of hundreds/thousands of robots, with promising applications to automated manufacturing, disaster relief, harvesting, last-mile delivery, port/airport operations, or search and rescue. The community has leveraged model-free multi-agent reinforcement learning (MARL) to devise efficient, scalable controllers for multi-robot systems (MRS). This review aims to provide an analysis of the state-of-the-art in distributed MARL for multi-robot cooperation. Recent findings: Decentralized MRS face fundamental challenges, such as non-stationarity and partial observability. Building upon the "centralized training, decentralized execution" paradigm, recent MARL approaches include independent learning, centralized critic, value decomposition, and communication learning approaches. Cooperative behaviors are demonstrated through AI benchmarks and fundamental real-world robotic capabilities such as multi-robot motion/path planning. Summary: This survey reports the challenges surrounding decentralized model-free MARL for multi-robot cooperation and existing classes of approaches. We present benchmarks and robotic applications along with a discussion on current open avenues for research.

Harsh Goel, Yifeng Zhang, Mehul Damani et al.

Many recent works have turned to multi-agent reinforcement learning (MARL) for adaptive traffic signal control to optimize the travel time of vehicles over large urban networks. However, achieving effective and scalable cooperation among junctions (agents) remains an open challenge, as existing methods often rely on extensive, non-generalizable reward shaping or on non-scalable centralized learning. To address these problems, we propose a new MARL method for traffic signal control, SocialLight, which learns cooperative traffic control policies by distributedly estimating the individual marginal contribution of agents on their local neighborhood. SocialLight relies on the Asynchronous Actor Critic (A3C) framework, and makes learning scalable by learning a locally-centralized critic conditioned over the states and actions of neighboring agents, used by agents to estimate individual contributions by counterfactual reasoning. We further introduce important modifications to the advantage calculation that help stabilize policy updates. These modifications decouple the impact of the neighbors' actions on the computed advantages, thereby reducing the variance in the gradient updates. We benchmark our trained network against state-of-the-art traffic signal control methods on standard benchmarks in two traffic simulators, SUMO and CityFlow. Our results show that SocialLight exhibits improved scalability to larger road networks and better performance across usual traffic metrics.

Jianghong Dong, Yifeng Zhang, Jiawei Wang et al.

Formation control of wheeled mobile robots (WMRs) has been extensively studied due to its broad applications in fields such as logistics transportation, environmental monitoring, and search and rescue. However, most existing works mainly focus on tracking predefined formations, which limits their adaptability to complex real-world environments. To address this, we propose REACT (Real-time Environment-Adaptive architecture for Continuous formation navigaTion), a hierarchical architecture integrating centralized formation generation and distributed formation maintenance. Specifically, our upper layer generates new environment-adaptive formations when necessary and uses our proposed TCF-R2T (Trajectory-Conflict-Free Robot-to-Target assignment) algorithm to compute conflict-free WMR-to-target assignments in polynomial time, enabling timely formation transitions without trajectory conflicts. At the lower layer, each WMR executes our developed JSTP (Joint Spatio-Temporal trajectory Planning) method to maintain the generated formation by simultaneously optimizing spatial positions and temporal durations, thereby enhancing coordination among WMRs and enabling continuous navigation in obstacle-rich environments and dynamic-obstacle scenarios. Both simulation and real-world experiments validate the effectiveness and practical applicability of REACT. Experimental videos are available on our project website: https://dongjh20.github.io/REACT-website.

Yifeng Zhang, Jieming Chen, Tingguang Zhou et al.

Multi-Agent Self-Driving (MASD) systems provide an effective solution for coordinating autonomous vehicles to reduce congestion and enhance both safety and operational efficiency in future intelligent transportation systems. Multi-Agent Reinforcement Learning (MARL) has emerged as a promising approach for developing advanced end-to-end MASD systems. However, achieving efficient and safe collaboration in dynamic MASD systems remains a significant challenge in dense scenarios with complex agent interactions. To address this challenge, we propose a novel collaborative(CO-) interaction-aware(-IN) MARL framework, named COIN. Specifically, we develop a new counterfactual individual-global twin delayed deep deterministic policy gradient (CIG-TD3) algorithm, crafted in a "centralized training, decentralized execution" (CTDE) manner, which aims to jointly optimize the individual objectives (navigation) and the global objectives (collaboration) of agents. We further introduce a dual-level interaction-aware centralized critic architecture that captures both local pairwise interactions and global system-level dependencies, enabling more accurate global value estimation and improved credit assignment for collaborative policy learning. We conduct extensive simulation experiments in dense urban traffic environments, which demonstrate that COIN consistently outperforms other advanced baseline methods in both safety and efficiency across various system sizes. These results highlight its superiority in complex and dynamic MASD scenarios, as further validated through real-world robot demonstrations. Supplementary videos are available at https://marmotlab.github.io/COIN/

Yifeng Zhang, Peizhuo Li, Tingguang Zhou et al.

Adaptive Traffic Signal Control (ATSC) aims to optimize traffic flow and minimize delays by adjusting traffic lights in real time. Recent advances in Multi-agent Reinforcement Learning (MARL) have shown promise for ATSC, yet existing approaches still suffer from limited representational capacity, often leading to suboptimal performance and poor generalization in complex and dynamic traffic environments. On the other hand, Large Language Models (LLMs) excel at semantic representation, reasoning, and analysis, yet their propensity for hallucination and slow inference speeds often hinder their direct application to decision-making tasks. To address these challenges, we propose a novel learning paradigm named LATS that integrates LLMs and MARL, leveraging the former's strong prior knowledge and inductive abilities to enhance the latter's decision-making process. Specifically, we introduce a plug-and-play teacher-student learning module, where a trained embedding LLM serves as a teacher to generate rich semantic features that capture each intersection's topology structures and traffic dynamics. A much simpler (student) neural network then learns to emulate these features through knowledge distillation in the latent space, enabling the final model to operate independently from the LLM for downstream use in the RL decision-making process. This integration significantly enhances the overall model's representational capacity across diverse traffic scenarios, thus leading to more efficient and generalizable control strategies. Extensive experiments across diverse traffic datasets empirically demonstrate that our method enhances the representation learning capability of RL models, thereby leading to improved overall performance and generalization over both traditional RL and LLM-only approaches. [...]

Mingliang Zhang, Sichang Su, Chengyang He et al.

In multi-agent reinforcement learning (MARL), achieving multi-task generalization to diverse agents and objectives presents significant challenges. Existing online MARL algorithms primarily focus on single-task performance, but their lack of multi-task generalization capabilities typically results in substantial computational waste and limited real-life applicability. Meanwhile, existing offline multi-task MARL approaches are heavily dependent on data quality, often resulting in poor performance on unseen tasks. In this paper, we introduce HyGen, a novel hybrid MARL framework, Hybrid Training for Enhanced Multi-Task Generalization, which integrates online and offline learning to ensure both multi-task generalization and training efficiency. Specifically, our framework extracts potential general skills from offline multi-task datasets. We then train policies to select the optimal skills under the centralized training and decentralized execution paradigm (CTDE). During this stage, we utilize a replay buffer that integrates both offline data and online interactions. We empirically demonstrate that our framework effectively extracts and refines general skills, yielding impressive generalization to unseen tasks. Comparative analyses on the StarCraft multi-agent challenge show that HyGen outperforms a wide range of existing solely online and offline methods.

Xibei Chen, Yifeng Zhang, Yuxiang Xiao et al.

Recent advances in robotics, automation, and artificial intelligence have enabled urban traffic systems to operate with increasing autonomy towards future smart cities, powered in part by the development of adaptive traffic signal control (ATSC), which dynamically optimizes signal phases to mitigate congestion and optimize traffic. However, achieving effective and generalizable large-scale ATSC remains a significant challenge due to the diverse intersection topologies and highly dynamic, complex traffic demand patterns across the network. Existing RL-based methods typically use a single shared policy for all scenarios, whose limited representational capacity makes it difficult to capture diverse traffic dynamics and generalize to unseen environments. To address these challenges, we propose CROSS, a novel Mixture-of-Experts (MoE)-based decentralized RL framework for generalizable ATSC. We first introduce a Predictive Contrastive Clustering (PCC) module that forecasts short-term state transitions to identify latent traffic patterns, followed by clustering and contrastive learning to enhance pattern-level representation. We further design a Scenario-Adaptive MoE module that augments a shared policy with multiple experts, thus enabling adaptive specialization and more flexible scenario-specific strategies. We conduct extensive experiments in the SUMO simulator on both synthetic and real-world traffic datasets. Compared with state-of-the-art baselines, CROSS achieves superior performance and generalization through improved representation of diverse traffic scenarios.

Fengxiaoxiao Li, Xiao Mao, Mingfeng Fan et al.

Robotic systems often require a team of robots to collectively visit multiple targets while optimizing competing objectives, such as total travel cost and makespan. This setting can be formulated as the Multi-Objective Multiple Traveling Salesman Problem (MOMTSP). Although learning-based methods have shown strong performance on the single-agent TSP and multi-objective TSP variants, they rarely address the combined challenges of multi-agent coordination and multi-objective trade-offs, which introduce dual sources of complexity. To bridge this gap, we propose CAMO, a conditional neural solver for MOMTSP that generalizes across varying numbers of targets, agents, and preference vectors, and yields high-quality approximations to the Pareto front (PF). Specifically, CAMO consists of a conditional encoder to fuse preferences into instance representations, enabling explicit control over multi-objective trade-offs, and a collaborative decoder that coordinates all agents by alternating agent selection and node selection to construct multi-agent tours autoregressively. To further improve generalization, we train CAMO with a REINFORCE-based objective over a mixed distribution of problem sizes. Extensive experiments show that CAMO outperforms both neural and conventional heuristics, achieving a closer approximation of PFs. In addition, ablation results validate the contributions of CAMO's key components, and real-world tests on a mobile robot platform demonstrate its practical applicability.

Jie Chen, Yuxin Cai, Yizhuo Wang et al.

Enabling robots to navigate open-world environments via natural language is critical for general-purpose autonomy. Yet, Vision-Language Navigation has relied on end-to-end policies trained on expensive, embodiment-specific robot data. While recent foundation models trained on vast simulation data show promise, the challenge of scaling and generalizing due to the limited scene diversity and visual fidelity in simulation persists. To address this gap, we propose ImagiNav, a novel modular paradigm that decouples visual planning from robot actuation, enabling the direct utilization of diverse in-the-wild navigation videos. Our framework operates as a hierarchy: a Vision-Language Model first decomposes instructions into textual subgoals; a finetuned generative video model then imagines the future video trajectory towards that subgoal; finally, an inverse dynamics model extracts the trajectory from the imagined video, which can then be tracked via a low-level controller. We additionally develop a scalable data pipeline of in-the-wild navigation videos auto-labeled via inverse dynamics and a pretrained Vision-Language Model. ImagiNav demonstrates strong zero-shot transfer to robot navigation without requiring robot demonstrations, paving the way for generalist robots that learn navigation directly from unlabeled, open-world data.

Shuhao Liao, Weihang Xia, Yuhong Cao et al.

The Multi-Agent Path Finding (MAPF) problem aims to determine the shortest and collision-free paths for multiple agents in a known, potentially obstacle-ridden environment. It is the core challenge for robotic deployments in large-scale logistics and transportation. Decentralized learning-based approaches have shown great potential for addressing the MAPF problems, offering more reactive and scalable solutions. However, existing learning-based MAPF methods usually rely on agents making decisions based on a limited field of view (FOV), resulting in short-sighted policies and inefficient cooperation in complex scenarios. There, a critical challenge is to achieve consensus on potential movements between agents based on limited observations and communications. To tackle this challenge, we introduce a new framework that applies sheaf theory to decentralized deep reinforcement learning, enabling agents to learn geometric cross-dependencies between each other through local consensus and utilize them for tightly cooperative decision-making. In particular, sheaf theory provides a mathematical proof of conditions for achieving global consensus through local observation. Inspired by this, we incorporate a neural network to approximately model the consensus in latent space based on sheaf theory and train it through self-supervised learning. During the task, in addition to normal features for MAPF as in previous works, each agent distributedly reasons about a learned consensus feature, leading to efficient cooperation on pathfinding and collision avoidance. As a result, our proposed method demonstrates significant improvements over state-of-the-art learning-based MAPF planners, especially in relatively large and complex scenarios, demonstrating its superiority over baselines in various simulations and real-world robot experiments. The code is available at https://github.com/marmotlab/SIGMA

Apoorva Vashisth, Manav Kulshrestha, Pranav Bakshi et al.

Visual navigation typically assumes the existence of at least one obstacle-free path between start and goal, which must be discovered/planned by the robot. However, in real-world scenarios, such as home environments and warehouses, clutter can block all routes. Targeted at such cases, we introduce the Lifelong Interactive Navigation problem, where a mobile robot with manipulation abilities can move clutter to forge its own path to complete sequential object- placement tasks - each involving placing an given object (eg. Alarm clock, Pillow) onto a target object (eg. Dining table, Desk, Bed). To address this lifelong setting - where effects of environment changes accumulate and have long-term effects - we propose an LLM-driven, constraint-based planning framework with active perception. Our framework allows the LLM to reason over a structured scene graph of discovered objects and obstacles, deciding which object to move, where to place it, and where to look next to discover task-relevant information. This coupling of reasoning and active perception allows the agent to explore the regions expected to contribute to task completion rather than exhaustively mapping the environment. A standard motion planner then executes the corresponding navigate-pick-place, or detour sequence, ensuring reliable low-level control. Evaluated in physics-enabled ProcTHOR-10k simulator, our approach outperforms non-learning and learning-based baselines. We further demonstrate our approach qualitatively on real-world hardware.

Yutong Wang, Pengliang Ji, Chaoqun Yang et al.

The LLM-as-a-Judge paradigm shows promise for evaluating generative content but lacks reliability in reasoning-intensive scenarios, such as programming. Inspired by recent advances in reasoning models and shifts in scaling laws, we pioneer bringing test-time computation into LLM-as-a-Judge, proposing MCTS-Judge, a resource-efficient, System-2 thinking framework for code correctness evaluation. MCTS-Judge leverages Monte Carlo Tree Search (MCTS) to decompose problems into simpler, multi-perspective evaluations. Through a node-selection strategy that combines self-assessment based on historical actions in the current trajectory and the Upper Confidence Bound for Trees based on prior rollouts, MCTS-Judge balances global optimization and refinement of the current trajectory. We further designed a high-precision, unit-test-level reward mechanism to encourage the Large Language Model (LLM) to perform line-by-line analysis. Extensive experiments on three benchmarks and five LLMs demonstrate the effectiveness of MCTS-Judge, which improves the base model's accuracy from 41% to 80%, surpassing the o1-series models with 3x fewer tokens. Further evaluations validate the superiority of its reasoning trajectory in logic, analytics, thoroughness, and overall quality, while revealing the test-time scaling law of the LLM-as-a-Judge paradigm.

Peizhuo Li, Hongyi Li, Ge Sun et al.

Despite recent advances in learning-based controllers for legged robots, deployments in human-centric environments remain limited by safety concerns. Most of these approaches use position-based control, where policies output target joint angles that must be processed by a low-level controller (e.g., PD or impedance controllers) to compute joint torques. Although impressive results have been achieved in controlled real-world scenarios, these methods often struggle with compliance and adaptability when encountering environments or disturbances unseen during training, potentially resulting in extreme or unsafe behaviors. Inspired by how animals achieve smooth and adaptive movements by controlling muscle extension and contraction, torque-based policies offer a promising alternative by enabling precise and direct control of the actuators in torque space. In principle, this approach facilitates more effective interactions with the environment, resulting in safer and more adaptable behaviors. However, challenges such as a highly nonlinear state space and inefficient exploration during training have hindered their broader adoption. To address these limitations, we propose SATA, a bio-inspired framework that mimics key biomechanical principles and adaptive learning mechanisms observed in animal locomotion. Our approach effectively addresses the inherent challenges of learning torque-based policies by significantly improving early-stage exploration, leading to high-performance final policies. Remarkably, our method achieves zero-shot sim-to-real transfer. Our experimental results indicate that SATA demonstrates remarkable compliance and safety, even in challenging environments such as soft/slippery terrain or narrow passages, and under significant external disturbances, highlighting its potential for practical deployments in human-centric and safety-critical scenarios.

He Jiang, Yutong Wang, Rishi Veerapaneni et al.

Lifelong Multi-Agent Path Finding (LMAPF) repeatedly finds collision-free paths for multiple agents that are continually assigned new goals when they reach current ones. Recently, this field has embraced learning-based methods, which reactively generate single-step actions based on individual local observations. However, it is still challenging for them to match the performance of the best search-based algorithms, especially in large-scale settings. This work proposes an imitation-learning-based LMAPF solver that introduces a novel communication module as well as systematic single-step collision resolution and global guidance techniques. Our proposed solver, Scalable Imitation Learning for LMAPF (SILLM), inherits the fast reasoning speed of learning-based methods and the high solution quality of search-based methods with the help of modern GPUs. Across six large-scale maps with up to 10,000 agents and varying obstacle structures, SILLM surpasses the best learning- and search-based baselines, achieving average throughput improvements of 137.7% and 16.0%, respectively. Furthermore, SILLM also beats the winning solution of the 2023 League of Robot Runners, an international LMAPF competition. Finally, we validated SILLM with 10 real robots and 100 virtual robots in a mock warehouse environment.

Yutong Wang, Pengliang Ji, Kaixin Li et al.

Large Language Reasoning Models have demonstrated remarkable success on static tasks, yet their application to multi-round agentic planning in interactive environments faces two fundamental challenges. First, the intractable credit assignment problem renders conventional reinforcement learning ineffective in sparse-reward settings. Second, the computational overhead of verbose, step-by-step reasoning histories is prohibitive. To address these challenges, we propose BPO, a three-stage framework (bootstrapping, extrapolation, and refinement) that establishes a self-improving data flywheel to develop robust reasoning models for long-horizon, sparse-reward environments. Our framework first bootstraps efficient reasoning using the proposed planning quaternions with long-short chain-of-thought fusion. It then extrapolates to out-of-distribution tasks through complexity-stratified curriculum learning. Finally, the model iteratively refines itself by learning exclusively on experiences selected via reward-gated rejection sampling. Experiments on ALFWorld, ScienceWorld, and WebShop demonstrate that our approach achieves state-of-the-art with significant token efficiency, providing a new recipe for reasoning models in agentic planning.

Mingfeng Fan, Jiaqi Cheng, Yaoxin Wu et al.

In recent years, deep reinforcement learning (DRL) has gained traction for solving the NP-hard traveling salesman problem (TSP). However, limited attention has been given to the close-enough TSP (CETSP), primarily due to the challenge introduced by its neighborhood-based visitation criterion, wherein a node is considered visited if the agent enters a compact neighborhood around it. In this work, we formulate a Markov decision process (MDP) for CETSP using a discretization scheme and propose a novel unified dual-decoder DRL (UD3RL) framework that separates decision-making into node selection and waypoint determination. Specifically, an adapted encoder is employed for effective feature extraction, followed by a node-decoder and a loc-decoder to handle the two sub-tasks, respectively. A k-nearest neighbors subgraph interaction strategy is further introduced to enhance spatial reasoning during location decoding. Furthermore, we customize the REINFORCE algorithm to train UD3RL as a unified model capable of generalizing across different problem sizes and varying neighborhood radius types (i.e., constant and random radii). Experimental results show that UD3RL outperforms conventional methods in both solution quality and runtime, while exhibiting strong generalization across problem scales, spatial distributions, and radius ranges, as well as robustness to dynamic environments.

Yifeng Zhang, Yilin Liu, Ping Gong et al.

Adaptive traffic signal control (ATSC) is crucial in reducing congestion, maximizing throughput, and improving mobility in rapidly growing urban areas. Recent advancements in parameter-sharing multi-agent reinforcement learning (MARL) have greatly enhanced the scalable and adaptive optimization of complex, dynamic flows in large-scale homogeneous networks. However, the inherent heterogeneity of real-world traffic networks, with their varied intersection topologies and interaction dynamics, poses substantial challenges to achieving scalable and effective ATSC across different traffic scenarios. To address these challenges, we present Unicorn, a universal and collaborative MARL framework designed for efficient and adaptable network-wide ATSC. Specifically, we first propose a unified approach to map the states and actions of intersections with varying topologies into a common structure based on traffic movements. Next, we design a Universal Traffic Representation (UTR) module with a decoder-only network for general feature extraction, enhancing the model's adaptability to diverse traffic scenarios. Additionally, we incorporate an Intersection Specifics Representation (ISR) module, designed to identify key latent vectors that represent the unique intersection's topology and traffic dynamics through variational inference techniques. To further refine these latent representations, we employ a contrastive learning approach in a self-supervised manner, which enables better differentiation of intersection-specific features. Moreover, we integrate the state-action dependencies of neighboring agents into policy optimization, which effectively captures dynamic agent interactions and facilitates efficient regional collaboration. Our results show that Unicorn outperforms other methods across various evaluation metrics, highlighting its potential in complex, dynamic traffic networks.

Yutong Wang, Guillaume Sartoretti

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.

Baxi Chong, Yasemin O. Aydin, Jennifer M. Rieser et al.

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.

Yuhong Cao, Zhanhong Sun, Guillaume Sartoretti

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 among all agents. Many robotic deployments require recomputing potentially large mTSP instances frequently, making the natural trade-off between computing time and solution quality of great importance. However, exact and heuristic algorithms become inefficient as the number of cities increases, due to their computational complexity. Encouraged by the recent developments in deep reinforcement learning (dRL), this work approaches the mTSP as a cooperative task and introduces DAN, a decentralized attention-based neural method that aims at tackling this key trade-off. In DAN, agents learn fully decentralized policies to collaboratively construct a tour, by predicting each other's future decisions. Our model relies on the Transformer architecture and is trained using multi-agent RL with parameter sharing, providing natural scalability to the numbers of agents and cities. Our experimental results on small- to large-scale mTSP instances ($50$ to $1000$ cities and $5$ to $20$ agents) show that DAN is able to match or outperform state-of-the-art solvers while keeping planning times low. In particular, given the same computation time budget, DAN outperforms all conventional and dRL-based baselines on larger-scale instances (more than 100 cities, more than 5 agents), and exhibits enhanced agent collaboration. A video explaining our approach and presenting our results is available at \url{https://youtu.be/xi3cLsDsLvs}.

Florian Laurent, Manuel Schneider, Christian Scheller et al.

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.

Sharada Mohanty, Erik Nygren, Florian Laurent et al.

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.

Mehul Damani, Zhiyao Luo, Emerson Wenzel et al.

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 warehouse 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 enable 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 and compare its performance to state-of-the-art planners in terms of makespan and throughput. We show that PRIMAL2 significantly surpasses our previous work and performs comparably to these baselines, while allowing real-time re-planning and scaling up to 2048 agents.

Samuel Shaw, Emerson Wenzel, Alexis Walker et al.

Multi-agent foraging (MAF) involves distributing a team of agents to search an environment and extract resources from it. Nature provides several examples of highly effective foragers, where individuals within the foraging collective use biological markers (e.g., pheromones) to communicate critical information to others via the environment. In this work, we propose ForMIC, a distributed reinforcement learning MAF approach that endows agents with implicit communication abilities via their shared environment. However, learning efficient policies with stigmergic interactions is highly nontrivial, since agents need to perform well to send each other useful signals, but also need to sense others' signals to perform well. In this work, we develop several key learning techniques for training policies with stigmergic interactions, where such a circular dependency is present. By relying on clever curriculum learning design, action filtering, and the introduction of non-learning agents to increase the agent density at training time at low computational cost, we develop a minimal learning framework that leads to the stable training of efficient stigmergic policies. We present simulation results which demonstrate that our learned policy outperforms existing state-of-the-art MAF algorithms in a set of experiments that vary team size, number and placement of resources, and key environmental dynamics not seen at training time.

Guillaume Sartoretti, William Paivine, Yunfei Shi et al.

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 accepted for publication as a regular paper in May 2019.

Guillaume Sartoretti, Justin Kerr, Yunfei Shi et al.

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.

Hadi Salman, Puneet Singhal, Tanmay Shankar et al.

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.