SHIFT Planner: Speedy Hybrid Iterative Field and Segmented Trajectory Optimization with IKD-tree for Uniform Lightweight Coverage

IROS 2025

Zexuan Fan1, Sunchun Zhou1, Junyi Cai1, Hengye Yang1, Ran Cheng1, Lige Liu1, Tao Sun1,
1Robot Zone. Midea Group Corporate
Paper arXiv Code Video Poster

Abstract

Efficient and uniform coverage of complex environments is critical for autonomous robots performing tasks such as cleaning, inspection, and agricultural operations. Existing methods often focus on minimizing traversal time or path length, neglecting the importance of uniform coverage and environmental attribute modeling. This paper introduces SHIFT Planner, a comprehensive planning and navigation framework tailored for these robotic tasks. Key innovations include:

  • Differential Geometry-Based Surface Extraction:
    • Extending uniform coverage planning from 2D to 3D environments using actual terrain point clouds collected from sensors like LiDAR and cameras.
  • Direct Landmark Point Planning:
    • Eliminating the need for region decomposition by deploying uniform coverage patterns directly onto environmental surfaces.
  • Dynamic Trajectory Planning with Safety Thresholds:
    • Introducing a local optimization framework that activates only when necessary, enhancing real-time performance and computational efficiency.
  • Environmental Attribute Modeling with Gaussian Fields
    • Adjusting robot behavior dynamically based on environmental attributes such as dirtiness or dryness.
  • Integration of IKD-tree for Lightweight Mapping
    • Utilizing an Incremental K-Dimensional tree (IKD-tree) for efficient local mapping and optimization.

Video

3D Optimization Process

Results on 3D path

Results on 2D path

Illustration of the SHIFT Planner path optimization. Given the original path, continuously iterate and optimize the trajectory. The red dashed line represents the initial path, blue represents the optimized path, and green represents the smooth path

Semantic map

Result on zigzag path

Result on 3D terrain speed adaptive coverage path

Result on 2D zigzag speed adaptive coverage path

Conference Poster

BibTeX

@inproceedings{frank2025shift,
  title     = {{SHIFT Planner: Speedy Hybrid Iterative Field and Segmented Trajectory Optimization with IKD-tree for Uniform Lightweight Coverage}},
  author    = {Zexuan Fan, Sunchun Zhou, Junyi, Cai},
  booktitle = {IEEE/RSJ International Conference on Intelligent Robots and Systems (IROS)},
  year      = {2025}
}