Semantic and Terrain-Aware Trajectory Optimization for Uniform Coverage in Obstacle-Laden Environments
Zexuan Fan, Hengye Yang, Sunchun Zhou, Junyi Cai, Tao Sun, Chang Liu
AI summary
Problem
Existing coverage path planning algorithms often neglect uniform coverage quality, 3D terrain adaptability, semantic cost integration, and real-time dynamic obstacle avoidance, leading to suboptimal performance in unstructured environments.
Approach
The SHIFT planner integrates a Gaussian diffusion-based speed adjustment for semantic/terrain-aware coverage (RFICP) with an incremental KD-tree guided sliding window optimization (IKD-SWOpt) for real-time, low-overhead local trajectory refinement.
Key results
- RFICP dynamically adjusts robot speed based on semantic intensity and 3D terrain curvature
- IKD-SWOpt reduces per-iteration planning time to 1.6 ms (0.9 ms with GPU) while maintaining clearance
- State-of-the-art coverage uniformity and adaptability across simulated drone and real-world vacuum experiments
- Eliminates ESDF construction overhead while matching benchmarks in flight time and path length
Why it matters
Enables autonomous robots in cleaning, inspection, and agriculture to perform thorough, terrain-adaptive coverage efficiently in dynamic, obstacle-rich real-world settings.
Abstract
Achieving efficient and uniform coverage in obstacle-laden unknown environments is essential for au- tonomous robots in cleaning, inspection and agricultural op- erations. Unlike most existing approaches that prioritize path length and time optimality, we propose the SHIFT planner framework, which integrates semantic mapping, adaptive cov- erage planning, and real-time obstacle avoidance to ensure comprehensive coverage across diverse terrains and seman- tic features. We first develop an innovative Radiant-Field- Informed Coverage Planning (RFICP) algorithm, which gen- erates trajectories that adapt to terrain variations. A Gaussian diffusion field is employed to adaptively adjust the robot’s speed, ensuring efficient coverage under varying environmental conditions influenced by semantic attributes. Next, we present a novel incremental KD-tree sliding window optimization (IKD- SWOpt) method to effectively handle unforeseen obstacles. IKD-SWOpt leverages an enhanced A* algorithm guided by the IKD-tree distance field to generate initial local avoidance tra- jectories. Subsequently, it optimizes trajectory segments within and outside waypoint safety zones by evaluating and refining non-compliant segments using an adaptive sliding window. This method not only reduces computational overhead but also guarantees high-quality real-time obstacle avoidance. Extensive experiments were conducted using drones in simulated envi- ronments and robotic vacuum cleaners in real-world settings. The SHIFT planner demonstrates state-of-the-art performance in coverage uniformity and adaptability across various terrains while maintaining very low computational overhead.