adaptive decoupling planning

Multi-robot motion planning faces significant challenges due to the high dimensionality of the joint configuration space, which incurs substantial computational costs and coordination difficulties. This work proposes an iteratively refined workspace decomposition approach that enables efficient cooperative path planning by hierarchically expanding subproblems and performing discrete search within decoupled, low-dimensional configuration spaces, thereby avoiding explicit construction of the high-dimensional joint space. By preserving coordination guarantees while substantially reducing computational complexity, the method achieves up to an order-of-magnitude improvement in planning speed compared to existing techniques, significantly enhancing the scalability and real-time performance of multi-robot systems.

This study addresses the computational bottleneck of global integer linear programming (ILP) when eliminating redundant moves during post-optimization in multi-agent path planning. To overcome this, we propose an exact decomposition mechanism that partitions problem instances into independent subproblems. For the majority of scenarios requiring no coordination, a lightweight single-agent solver is designed; for those necessitating coordination, a hybrid planning algorithm combining CBS-style search with region awareness is employed, comprehensively replacing conventional commercial ILP solvers. While preserving identical solution quality, our approach achieves a median 10.5× speedup per instance. Notably, the lightweight solving component operates approximately 1900 times faster than Judgelight, substantially improving the computational efficiency of post-optimization.

This paper addresses autonomous planning for critical agents in multi-agent environments where adversaries’ behaviors are unknown. Method: We propose the first unified theoretical framework that systematically characterizes the dynamic trade-off between information exploitation and exploration. The framework encompasses a spectrum of typed planners—from exact to approximate—and formally instantiates “safe agents” and their knowledge-augmented variants as special cases. Our approach integrates typed planning, Bayesian inference, online replanning, and multi-agent path modeling, yielding an engineering implementation of 13 distinct planners. Contribution/Results: Extensive experiments on path planning tasks with up to 50 agents validate a performance gradient across planners. Notably, safe-agent–based methods achieve robust and efficient decision-making across most scenarios with low computational overhead, significantly enhancing practicality and scalability in unknown environments.

This study addresses the problem of collision-free cooperative path planning for multiple robots operating on graphs derived from the discretization of simple polygons, with the objective of minimizing the total travel distance across all robots. The work proposes a fixed-parameter tractable (FPT) algorithm parameterized by the number of robots \(k\), thereby establishing—for the first time—that this problem is FPT on such polygon-induced graphs. By leveraging structural properties of the underlying graph and geometric constraints inherent to simple polygons, the approach extends existing FPT results from grid environments to more general planar, obstacle-constrained settings. This advancement provides significant progress toward resolving open questions concerning subgrid and planar graph variants of multi-robot path planning.

This work addresses decentralized multi-agent navigation in cluttered environments, proposing the first joint optimization framework for agent policies and reconfigurable environmental layouts (e.g., obstacle placements). Methodologically, it employs model-free policy gradient reinforcement learning and introduces a two-stage alternating optimization algorithm that concurrently updates distributed agent policies and environmental structure. Theoretical analysis establishes convergence to local minima of a time-varying non-convex optimization problem. A key finding is that the optimized environment autonomously forms implicit, motion-decoupled guidance structures—enhancing behavioral coordination without explicit communication or centralized control. Experiments across diverse dense scenarios demonstrate consistent superiority over baselines in navigation success rate, throughput efficiency, and collision rate, empirically validating that environmental configuration optimization delivers substantial gains for multi-agent collaborative navigation.

This study addresses the problem of coordinating multiple robots on graph-structured environments—including grids, planar graphs, and unit disk graphs—to achieve a connected target formation without collisions while minimizing the total travel distance. To this end, the work proposes a unified framework that integrates movement minimization with multi-agent path planning, innovatively incorporating connectivity constraints into the formulation. The resulting model is solved by leveraging techniques from parameterized complexity theory and graph theory. The primary contribution lies in establishing tight computational complexity boundaries for this problem across various graph topologies. By rigorously delineating these theoretical limits, this research provides a solid foundation for the efficient coordination of multi-robot systems operating under spatial and connectivity constraints.

This work addresses the challenge of simultaneously achieving path optimality and kinematic feasibility in multi-agent motion planning by proposing a two-stage framework. It first generates initial collision-free paths using Conflict-Based Search (CBS) or Priority-Based Search (PBS), then refines them through a multi-phase optimal control problem (OCP) formulation. A motion primitive generation mechanism is introduced to enforce uniform sampling time constraints. Additionally, the SIPP-IP algorithm is extended to accommodate general cost functions and large-footprint agents. Experimental results demonstrate that, in trailer-truck systems, lattice-based planners outperform the extended SIPP-IP due to less conservative collision checking. In cluttered environments, CBS yields higher success rates and lower computation times than PBS, though both approaches produce solutions of comparable quality after optimization.

This study addresses a novel multi-agent path planning problem in which multiple agents originate from a common source and must cooperatively navigate toward a shared destination through a graph containing periodically cooling hazardous nodes; any agent contacting such a node is eliminated. The objective is to maximize the number of agents that successfully reach the goal. Within a deterministic, fully observable discrete-time setting, the problem is formally defined for the first time. The authors establish that the length of an optimal solution is polynomially bounded and prove that the problem remains NP-hard even when restricted to tree-structured graphs. However, they also identify a tractable case: when the underlying graph consists of vertex-disjoint paths, the problem admits a polynomial-time algorithm. Through rigorous complexity analysis and graph-theoretic modeling, this work delineates the computational boundaries of the problem.

Multi-robot motion planning faces the challenge of simultaneously achieving fast initial solution generation, efficient optimization of solution quality, and scalability. This work proposes AO-ARC, the first approach to integrate the anytime-optimal AO-x meta-algorithm with an adaptive (de)coupled ARC solver. AO-ARC rapidly produces feasible solutions while guaranteeing asymptotic optimality with respect to makespan and maintaining consistent cost bounds across varying robot decompositions. Experimental results demonstrate that AO-ARC matches state-of-the-art feasibility solvers in initial solution speed and significantly outperforms existing anytime methods in both convergence rate of solution quality and reliability, across 2D coordination scenarios and 3D robotic arm tasks.

为解决多智能体系统在执行复杂时序任务时的可扩展性和泛化性问题，提出了一种基于扩散模型和信号时序逻辑的方法，增强规划多样性并减少碰撞。