Neural-Primitive: An Efficient End-to-End Local Planner With Primitive-Based Imitation Learning for Autonomous Flight
A lightweight offline-primitive-based dataset collection framework is designed to produce safe and high-quality trajectory primitives in non-convex environments and demonstrates superiority in both planning latency and target-reaching progress quality.
Zhi-Tao Liu, Guang-Tong Xu, Zi-Han Wang et al.
· IEEE Transactions on Industr... · 0 citations