LLM agents coordinate robot swarms: correction fixes collisions, teams search 2x faster
peter_richtarik · x · 2026-10-22
- peterrichtarik shares a multi-agent robot planning experiment: 3 GPT-OSS-120B agents route 9 robots on a 16×16 grid.
- With a correction mechanism, all 40 team runs were collision-free; without it, only 2 of 20 were — agents kept crashing into each other's robots.
- On speed, a two-floor task (18 robots, 6 agents) had median search time of 595s vs 1200s for one agent (2x faster); merging the best verified combination of proposals instead of picking one color cut mean search time by another 29–36% (chart: 29%, 36%, 31% gains across three maps).
More from Embodied
- RAX Uses a Frozen Coding Agent to Drive Robots: 96.7% on DexJoCo at 1/20 the Cost — du_yilun · 2026-10-22
- CMU to Launch Hands-On Humanoid Learning Course in Spring 2027, Fully Open-Sourced — GuanyaShi · 2026-10-22
- S1 data glove captures demos of Rubik's Cube solves, card shuffling, butterfly knife tricks — Haoyu_Xiong_ · 2026-10-22
- Recursia hits 28-53% zero-shot on LIBERO-Pro Long, far ahead of ASPIRE and π0.5 — jiqizhixin · 2026-10-22
- Tesla flips California Optimus plant status to "early ramp" as Texas site stays under construction — BrettKrieger12 · 2026-10-22
- S1 skill-capture glove records joint motion, touch and video for robot learning — ybisk · 2026-10-22