FELP: Fast and Effective Autonomous Flight on Large-Scale and Cluttered Environments Based on Unified Linear Parametric Map
hongyu nie, Xingyu Li, Xu Liu, Decai Li, Yuqing He
AI summary
Problem
AAVs face severe computational and storage bottlenecks when processing vast data in large-scale, cluttered environments. Additionally, traditional gradient-based planning methods rely on fixed resolutions for occupancy and ESDF maps, limiting adaptability and efficiency.
Approach
FELP projects irregular point clouds into a high-dimensional space using a Random Mapping Method to learn unified linear parametric models for both occupancy grids and ESDFs. This enables closed-form, resolution-independent distance calculations and accelerates front-end path searching and back-end trajectory optimization.
Key results
- Reduces mapping time by 68% compared to EGO-Planner
- Reduces planning time by 29% compared to EGO-Planner
- Enables continuous, memory-efficient occupancy grid mapping via a learned linear classifier
- Provides a closed-form, arbitrary-resolution ESDF map for real-time trajectory optimization
Why it matters
It provides a lightweight, scalable mapping and planning solution for resource-constrained aerial vehicles operating in complex, large-scale environments.
Abstract
Current AAV autonomous flights exhibit efficient per- formance in both indoor and field environments. However, they often face significant challenges in large-scale and cluttered en- vironments, where the vast amount of captured data can lead to computation and storage bottlenecks. Additionally, the existing gradient-based planning methods depend on appropriate resolu- tions to adapt to different scenarios. In this letter, we present FELP, a fast and effective autonomous flight system for large-scale and cluttered environments based on the unified linear paramet- ric map. It can enhance the adaptability of planners to diverse environments. First, by the random mapping method (RMM), the original irregular points in low-dimensional space are mapped into the high-dimensional space, where the points are approximately linearly separable or distributed. Leveraging the features of this mapping space, we can quickly obtain the occupancy state and Euclidean distance (the distance to the nearest obstacle) rather than relying on a large number of queries and repeated iterations. Then we learn a unified linear parametric model about grid maps and ESDF maps. Based on the linear parametric model, path searching is quickly executed in the front-end. Unlike traditional methods that compute the ESDF through interpolation, the closed-form ESDF can be solved efficiently, enabling real-time online trajectory optimization in the back-end. Compared to EGO-Planner, FELP reduces the mapping time by 68% and the planning time by 29%. Simulation and real-world experiments are conducted to verify their comprehensive performance compared to typical methods and state-of-the-art methods.