GRACE: A Unified 2D Multi-Robot Path Planning Simulator & Benchmark for Grid, Roadmap, And Continuous Environments

Chuanlong Zang; Anna Mannucci; Isabelle Barz; Philipp Schillinger; Florian Lier; Wolfgang Hönig

cs.ROMar 11, 2026

GRACE: A Unified 2D Multi-Robot Path Planning Simulator & Benchmark for Grid, Roadmap, And Continuous Environments

Authors: Chuanlong Zang, Anna Mannucci, Isabelle Barz, Philipp Schillinger, Florian Lier, Wolfgang Hönig

Organizations: Robert Bosch GmbH, Corporate Research, Stuttgart, Germany. · Technical University of Berlin, Berlin, Germany.

Abstract

Advancing Multi-Agent Pathfinding (MAPF) and Multi-Robot Motion Planning (MRMP) requires platforms that enable transparent, reproducible comparisons across modeling choices. Existing tools either scale under simplifying assumptions (grids, homogeneous agents) or offer higher fidelity with less comparable instrumentation. We present GRACE, a unified 2D simulator+benchmark that instantiates the same task at multiple abstraction levels (grid, roadmap, continuous) via explicit, reproducible operators and a common evaluation protocol. Our empirical results on public maps and representative planners enable commensurate comparisons on a shared instance set. Furthermore, we quantify the expected representation-fidelity trade-offs (MRMP solves instances at higher fidelity but lower speed, while grid/roadmap planners scale farther). By consolidating representation, execution, and evaluation, GRACE thereby aims to make cross-representation studies more comparable and provides a means to advance multi-robot planning research and its translation to practice.

Figures & tables

Panel of Figure 1 — Fig. 1: GRACE: Single scenario, multiple planning representations. Identical agents/scenarios, distinct trajectories from representation-specific constraints: (a) occupancy-grid (4-connected discrete actions), (b) graph abstraction (single integrator), and (c) continuous space (double integrator, heterogeneous footprint). Individual path lengths highlight these representation-specific constraints.

Platform	Rep.	CST Env.	CST Plan.	Dyn.	Het.	Scale
MovingAI-related [ 8 , 9 , 10 , 11 ]	G	✗	✗	✗	✗	1k
asprilo [ 12 ]	G	✗	✗	✗	✗	50
Flatland [ 13 ]	G	✗	✗	✗	✗	6k
LoRR [ 14 ]	G	✗	✗	✗	✗	10k
POGEMA [ 6 ]	G	✗	✗	✗	✗	1M
MRP-Bench [ 7 ]	C	✓	✗	✗	✗	10

TABLE I: Comparison of 2D multi-agent path planning platforms.

Planners	Rep.	Dyn.	Het.	Opt. †	Comp.	OSS	Eval.
CBSH2-RTC [ 18 ]	G	✗	✗	opt.	✓	✓	✓
EECBS [ 19 ]	G	✗	✗	subopt.	✓	✓	✓
LaCAM [ 2 ]	G	✗	✗	subopt.	✓	✓	✓
MAPF-LNS2 [ 20 ]	G	✗	✗	subopt.	✓	✓	✓
SCRIMP [ 21 ]	G	✗	✗	subopt.	✓	✓	✓
MAPF-GPT-DDG [ 22 ]	G	✗	✗	subopt.	✓	✓	✓

TABLE II: Comparison of 2D multi-agent path planning approaches.

Aspect	Continuous	Roadmap	Grid
Time	continuous	continuous	discretized
Footprint	$S_{i}$ (convex)	disc $r_{\mathrm{com}}$	disc $r_{\mathrm{com}}$
Orient.	if $Q_{i}{=}\mathrm{SE}(2)$	–	–
Kinematics	$f_{i}$	SI( $\nu$ )	unit moves
Collision	swept $S_{i}$	cap/edges/ints	MAPF rules
Goals	invariant set	remain at node	remain at cell

TABLE III: Feature mapping.

Fig. 3: Cross-Environment Comparison of MRMP and MAPF Solvers: SoC, Makespan, Planning Time (db-CBS=100%)

Fig. 4: Grid MAPF Comparison: Success Rate with Number of Agents.

Planner	Mean SoC (m) $\downarrow$	Mean planning time (s) $\downarrow$
CBSH2–RTC	3827 $\pm$ 286	4.61 $\pm$ 1.84
EECBS	3940 $\pm$ 288	0.63 $\pm$ 0.08
MAPF–LNS2	4206 $\pm$ 321	0.45 $\pm$ 0.05
MAPF–GPT–DDG	4257 $\pm$ 333	5.66 $\pm$ 0.26 †
LaCAM	4511 $\pm$ 358	0.01 $\pm$ 0.00
SCRIMP	5290 $\pm$ 422	20.86 $\pm$ 1.56 †

TABLE IV: SoC and planning time averaged over the common-success subset across all planners.

$N$	RTF (median)	Peak CPU (%)	Peak Mem (GB)
2	1288.21	9.9	0.039
550	10.59	108.6	0.134
1000	5.41	107.4	0.213
2000	2.39	107.1	0.387
2500	1.84	107.2	0.476

TABLE V: Key points across

N

during path-playback on grid representation warehouse_large using LaCAM.

Explore similar work

From Discrete Plans to Real-World Execution: A World-Model-Driven Framework for Execution-Aware Multi-Agent Path Finding
Nov 26, 2025Jingtian Yan, Shuai Zhou, He Jiang +2Multi-Robot Motion Planning World Models for Robotics
This paper was produced by the IEEE Publication Technology Group. They are in Piscataway, NJ.
Decoupling Geometric Planning and Execution in Scalable Multi-Agent Path Finding
Mar 11, 2026Fernando Salanova, Eduardo Montijano, Cristian MahuleaMulti-Agent Planning Multi-Agent Path Finding
Dept. of Systems Engineering and Computer Science, University of Zaragoza, Zaragoza, Spain
The Cost of Classical Multi-Agent Path Finding
Oct 7, 2026Alvin Combrink, Sabino Francesco Roselli, Martin FabianMulti-Agent Path Finding Graph-Based Planning
Chalmers University of Technology Gothenburg, Sweden

Nov 26, 2025J/K move · Enter open · S save

From Discrete Plans to Real-World Execution: A World-Model-Driven Framework for Execution-Aware Multi-Agent Path Finding

Jingtian Yan, Shuai Zhou, He Jiang, Stephen F. Smith+1

This paper was produced by the IEEE Publication Technology Group. They are in Piscataway, NJ.

Multi-Robot Motion Planning World Models for Robotics World Model-Based Planning Multi-Agent Path Finding

Open paper AskPDF arXiv

Multi-Agent Path Finding (MAPF) studies how to coordinate multiple agents to reach their goals without collisions and underpins a range of large-scale robotic systems, including automated warehousing and manufacturing. Recent advances enable MAPF solvers to compute high-quality plans for hundreds of agents. However, these plans are generated using simplified robot models with discretized time and action spaces. When they are deployed in physical systems, heterogeneous robot dynamics, asynchronous interactions, communication delays, and other real-world factors can lead to substantial deviations from planned performance. We bridge the gap between discrete planning and real-world execution through ExecTimeNet, a learned world model of MAPF execution that predicts how a discrete MAPF solution will unfold on physical robots, mapping each discrete action to its realized execution state, including its wall-clock completion time and the kinodynamic state in which it ends. Building on this capability, we first propose REMAP, an execution-aware MAPF framework that integrates execution-time estimation into planning, guiding the search toward MAPF solutions with improved execution performance. We also introduce ESADG, a post-planning optimization procedure that optimizes the execution schedule of a given MAPF solution while preserving path feasibility. We evaluate proposed frameworks in high-fidelity simulation with up to 300 agents and on physical robots. In simulation, ExecTimeNet predicts the execution state accurately and transfers to unseen maps and agent counts. Across simulation benchmarks spanning diverse map topologies, REMAP reduces delays by up to 21% over baselines, while ESADG achieves up to 40% normalized improvement. On physical hardware, the full pipeline reduces total execution time by up to 15.3%, demonstrating effective transfer from simulation to real-world deployment.

Abstract

Explore similar work

From Discrete Plans to Real-World Execution: A World-Model-Driven Framework for Execution-Aware Multi-Agent Path Finding

Decoupling Geometric Planning and Execution in Scalable Multi-Agent Path Finding

The Cost of Classical Multi-Agent Path Finding