FASTFUNNELS: Agile Multi Robot Navigation in Narrow Spaces via Adaptive Dynamic Funnels
FASTFUNNELS: Agile Multi Robot Navigation in Narrow Spaces via Adaptive Dynamic Funnels
This paper introduces FastFunnels, a method for high-speed multi-robot navigation in cluttered environments. Unlike rigid formation control that requires predefined shapes and slows down near bottlenecks, FastFunnels uses a reinforcement learning policy to generate a dynamic, blob-like elliptical patch that adapts its shape in real time. At the low level, distributed model predictive control (DMPC) keeps robots within this soft forcing function while avoiding collisions. Evaluated on simulated F1TENTH racing cars across 14 scenarios with 1 to 6 robots, FastFunnels achieves an 89% completion rate at an average speed of 7.73 m/s, significantly outperforming baselines that either fail or operate at much lower speeds (0.07-1.79 m/s) with only about 40% completion.
The authors present FastFunnels, a hierarchical algorithm combining reinforcement learning for dynamic shape adaptation and distributed model predictive control for collision avoidance, evaluated on simulated F1TENTH racing cars achieving an 89% completion rate at 7.73 m/s across 14 scenarios.
Source: arXiv Robotics — research abstracts · Read original article ↗
Loading article text…
What the source reports
Publisher-reported claims, with original evidence. These results have not been independently verified by RoboSignal.
- Environment
- SimulationOpen source S5
Reported numbers
Robot count
6 robots
View original evidence
nd orienting itself around obstacles and through tight spaces. At the low level, the robots are controlled via distributed model predictive control (DMPC) to stay inside the patch, which acts as a soft forcing function, while avoiding collisions with each other. We evaluate our approach, FastFunnels, on simulated F1TENTH racing cars across 14 scenarios with N=1 to N=6 robots, averaged over 20 trials each. FastFunnels
Open source S5trials
>20 trials
Reported trials
View original evidence
nd orienting itself around obstacles and through tight spaces. At the low level, the robots are controlled via distributed model predictive control (DMPC) to stay inside the patch, which acts as a soft forcing function, while avoiding collisions with each other. We evaluate our approach, FastFunnels, on simulated F1TENTH racing cars across 14 scenarios with N=1 to N=6 robots, averaged over 20 trials each. FastFunnels
Open source S5Reported completion rate
89%
View original evidence
achieves an 89% completion rate at 7.73 m/s, while all baselines either fail to complete the track or succeed only at very low speeds (0.07-1.79 m/s, about 40% completion rate).
Open source S6speed
7.73
View original evidence
achieves an 89% completion rate at 7.73 m/s, while all baselines either fail to complete the track or succeed only at very low speeds (0.07-1.79 m/s, about 40% completion rate).
Open source S6speed
0.07–1.79
View original evidence
achieves an 89% completion rate at 7.73 m/s, while all baselines either fail to complete the track or succeed only at very low speeds (0.07-1.79 m/s, about 40% completion rate).
Open source S6Reported completion rate
≈40%
View original evidence
achieves an 89% completion rate at 7.73 m/s, while all baselines either fail to complete the track or succeed only at very low speeds (0.07-1.79 m/s, about 40% completion rate).
Open source S6
- paper: not_reported
What remains unknown
Not established in the collected evidence: Control, Data origin.
Reported performance applies to the described task. It does not establish general autonomy or deployment readiness.
Source excerpts and review record
Automatically extracted; no manual editorial approval recorded.
nd orienting itself around obstacles and through tight spaces. At the low level, the robots are controlled via distributed model predictive control (DMPC) to stay inside the patch, which acts as a soft forcing function, while avoiding collisions with each other. We evaluate our approach, FastFunnels, on simulated F1TENTH racing cars across 14 scenarios with N=1 to N=6 robots, averaged over 20 trials each. FastFunnels
Open source S5
achieves an 89% completion rate at 7.73 m/s, while all baselines either fail to complete the track or succeed only at very low speeds (0.07-1.79 m/s, about 40% completion rate).
Open source S6
Source：arXiv Robotics — research abstracts · arxiv.org