Khái niệm cốt lõi
PRIMER leverages imitation learning to achieve near-optimal and computationally efficient multiagent trajectory planning for robots with limited perception in uncertain environments, addressing the limitations of traditional optimization-based methods.
Thống kê
PRIMER achieves a 5614-time reduction in computation time compared to PARM* in a single-agent, single-obstacle environment.
PRIMER maintains a 100% success rate and 0% dynamic constraint violations in the same environment.
In a multiagent and multi-obstacle environment with three agents and two obstacles, PRIMER achieves a high success rate, outperforming PARM* in terms of success rate and computation time.
Trích dẫn
"To tackle the challenges of (1) unknown objects detection and collision avoidance, (2) localization errors/uncertainties, (3) scalability, and (4) fast and efficient computation, we propose PRIMER, an IL-based decentralized, asynchronous, perception-aware multiagent trajectory planner."
"PRIMER leverages the low computational requirements at deployment of neural networks and achieves a computation speed up to 5614 times faster than optimization-based approaches."