MA-SafeDiffuser: Safe Multi-Agent Planning with Diffusion Probabilistic Models

We propose MA-SafeDiffuser, a multi-agent extension of SafeDiffuser that equips diffusion-based trajectory planners with finite-time diffusion invariance guarantees for joint-agent safety specifications. Building on the single-agent construction that embeds control barrier function (CBF) constraints into reverse-diffusion updates, we: (i) formalize joint safe sets as intersections of per-agent and pairwise barrier sets; (ii) derive centralized and decentralized (communication-aware) constrained denoising procedures with provable invariance under mild assumptions; (iii) address local-trap and deadlock phenomena via time-varying specifications and liveness CBFs; and (iv) develop a lightweight benchmarking suite including a multi-agent Maze2D domain. Empirically, MA-SafeDiffuser reduces violation counts relative to unconstrained diffusion baselines while retaining planning quality. Across 2, 4, and 8-agent centralized Maze2D settings, MA-SafeDiffuser reduces violations from 23.66 to 9.00, 42.66 to 11.33, and 87.33 to 40.00, respectively, while maintaining nearly unchanged task return. The decentralised formulation also reduces violations under local neighbour communication. With a learned MLP score denoiser in the four-agent setting, CBF enforcement further reduces violations from 125 to 62 while preserving successful task completion and comparable reward.