We showcase a range of projects in multi-agent systems, spanning from fundamental motion-planning methods for coordinating teams of robots to more involved challenges such as task allocation and integrated task-and-motion planning. Our work explores how agents can collaborate to complete shared objectives, navigate dynamic environments, and even influence the behavior or movement of other groups of agents. Several of these projects draw on a unified modeling approach that combines multi-agent simulation with roadmap-based planning, enabling us to address a broad spectrum of multi-agent coordination problems.

Projects
Multi-Agent Motion Planning

ARC: Adaptive Robot Coordination

Adaptive Robot Coordination warehouse scenario

A hybrid framework for multi-robot motion planning that uses local subproblems to resolve inter-robot conflicts, dynamically adjusting between coupled and decoupled planning to deliver efficient, probabilistically complete solutions.

Multi-Robot Motion Planning Hybrid Planning Conflict Resolution
View project
Conflict-Based Search

CBS Extensions to MR-TAMP

CBS extensions warehouse scenario

Hybrid Conflict-Based Search extensions, including CBS-MP and TMP-CBS, that scale multi-robot motion planning and task allocation to continuous, high-DOF, heterogeneous robot teams with complex inter-task dependencies.

CBS-MP TMP-CBS Task Allocation
View project
Task & Motion Planning

DaSH: Decomposable State Space Hypergraph

DaSH hypergraph overview

A hypergraph-based multi-robot task and motion planning algorithm that uses a decomposed state space representation, scaling linearly rather than exponentially with the number of robots and objects.

Hypergraph Planning MR-TAMP Decomposed State Space
View project
Workspace Guidance

Multi-Agent Workspace Guidance

Composite skeleton guidance

Extends topological skeleton guidance to multi-robot motion planning via CDR-RRT and Workspace-DaSH, leveraging workspace structure to efficiently coordinate teams of robots through narrow passages.

CDR-RRT Workspace-DaSH Topological Skeleton
View project
Multi-Robot Path Finding

Multi-Agent Path Finding

Multi-agent path finding

Hierarchical Composition Conflict-Based Search (HC-CBS) and its parallel extensions PHC-CBS and DPHC-CBS, which exploit the embarrassingly parallel structure of CBS to scale multi-agent pathfinding via multithreading.

HC-CBS Parallel Planning Hierarchical Search
View project
Modular Robots

Metamorphic Robots

Metamorphic modular robots

Decentralized algorithms for controlling collections of identical, independently controlled hexagonal modules that connect, disconnect, and reconfigure to navigate around irregular obstacles in dangerous environments.

Decentralized Control Modular Systems Reconfiguration
View project
Collision Avoidance

RRVO: Reciprocally-Rotating Velocity Obstacles

RRVO collision avoidance

An extension to ORCA that allows polygonal agents to actively rotate to avoid collisions, overcoming the deadlocking problem that arises when large agents with opposing goals meet in dense multi-agent simulations.

ORCA Extension Deadlock Resolution Multi-Agent Simulation
View project
Decentralized Planning

PRISM: Distributed Multi-Task Pathfinding

PRISM decentralized pathfinding

Pathfinding with Rapid Information Sharing using Motion Constraints — a decentralized MT-MAPF algorithm that uses information packets to share motion constraints, provably resolving deadlocks and scaling to large agent teams.

MT-MAPF Decentralized Deadlock-Free
View project
Interaction Planning

Robot Interaction

Interaction Template Method

The Interaction Template Method, a scalable motion-planning-first approach for multi-robot task and motion planning that embeds task semantics directly into a combined roadmap to coordinate heterogeneous robot teams.

Interaction Templates MR-TAMP Heterogeneous Teams
View project
Motion Planning

CIPHER: Scalable Multi-Robot Motion Planning

CIPHER workspace decomposition

Coordinated Incremental Planning with Hierarchical Expansion and Refinement — introduces resolution guidance that reasons over workspace representations at different resolutions to scale multi-robot motion planning.

Resolution Guidance Hierarchical Planning CBS
View project

Publications

Updated: