Logo
ScienceStack
Table of Contents
Fetching ...
Sign In

HE-Nav: A High-Performance and Efficient Navigation System for Aerial-Ground Robots in Cluttered Environments

Junming Wang, Zekai Sun, Xiuxian Guan, Tianxiang Shen, Dong Huang, Zongyuan Zhang, Tianyang Duan, Fangming Liu, Heming Cui

TL;DR

HE-Nav is presented, the first high-performance and efficient navigation system tailored for AGRs operating in cluttered environments and employs the energy-efficient kinodynamic A* search algorithm to guarantee planning is energy-saving.

Abstract

Existing AGR navigation systems have advanced in lightly occluded scenarios (e.g., buildings) by employing 3D semantic scene completion networks for voxel occupancy prediction and constructing Euclidean Signed Distance Field (ESDF) maps for collision-free path planning. However, these systems exhibit suboptimal performance and efficiency in cluttered environments with severe occlusions (e.g., dense forests or tall walls), due to limitations arising from perception networks' low prediction accuracy and path planners' high computational overhead. In this paper, we present HE-Nav, the first high-performance and efficient navigation system tailored for AGRs operating in cluttered environments. The perception module utilizes a lightweight semantic scene completion network (LBSCNet), guided by a bird's eye view (BEV) feature fusion and enhanced by an exquisitely designed SCB-Fusion module and attention mechanism. This enables real-time and efficient obstacle prediction in cluttered areas, generating a complete local map. Building upon this completed map, our novel AG-Planner employs the energy-efficient kinodynamic A* search algorithm to guarantee planning is energy-saving. Subsequent trajectory optimization processes yield safe, smooth, dynamically feasible and ESDF-free aerial-ground hybrid paths. Extensive experiments demonstrate that HE-Nav achieved 7x energy savings in real-world situations while maintaining planning success rates of 98% in simulation scenarios. Code and video are available on our project page: https://jmwang0117.github.io/HE-Nav/.

HE-Nav: A High-Performance and Efficient Navigation System for Aerial-Ground Robots in Cluttered Environments

TL;DR

HE-Nav is presented, the first high-performance and efficient navigation system tailored for AGRs operating in cluttered environments and employs the energy-efficient kinodynamic A* search algorithm to guarantee planning is energy-saving.

Abstract

Existing AGR navigation systems have advanced in lightly occluded scenarios (e.g., buildings) by employing 3D semantic scene completion networks for voxel occupancy prediction and constructing Euclidean Signed Distance Field (ESDF) maps for collision-free path planning. However, these systems exhibit suboptimal performance and efficiency in cluttered environments with severe occlusions (e.g., dense forests or tall walls), due to limitations arising from perception networks' low prediction accuracy and path planners' high computational overhead. In this paper, we present HE-Nav, the first high-performance and efficient navigation system tailored for AGRs operating in cluttered environments. The perception module utilizes a lightweight semantic scene completion network (LBSCNet), guided by a bird's eye view (BEV) feature fusion and enhanced by an exquisitely designed SCB-Fusion module and attention mechanism. This enables real-time and efficient obstacle prediction in cluttered areas, generating a complete local map. Building upon this completed map, our novel AG-Planner employs the energy-efficient kinodynamic A* search algorithm to guarantee planning is energy-saving. Subsequent trajectory optimization processes yield safe, smooth, dynamically feasible and ESDF-free aerial-ground hybrid paths. Extensive experiments demonstrate that HE-Nav achieved 7x energy savings in real-world situations while maintaining planning success rates of 98% in simulation scenarios. Code and video are available on our project page: https://jmwang0117.github.io/HE-Nav/.
Paper Structure (16 sections, 11 equations, 11 figures, 6 tables)

This paper contains 16 sections, 11 equations, 11 figures, 6 tables.

Table of Contents

  1. INTRODUCTION
  2. RELATED WORKS
  3. Motion Planning for AGRs
  4. Occlusion-Aware for AGRs
  5. Perception Module of HE-Nav
  6. LBSCNet Network Structure
  7. Aerial-Ground Motion Planner of HE-Nav
  8. Energy-Efficient Kinodynamic Hybrid A* Path Searching
  9. Gradient-Based B-spline Trajectory Optimization
  10. Evaluation
  11. Evaluation setup
  12. LBSCNet Comparison against the state-of-the-art.
  13. Simulated Air-Ground Robot Navigation
  14. Real-world Air-Ground Robot Navigation
  15. CONCLUSION
  16. ...and 1 more sections

Figures (11)

  • Figure 1: (a) Current navigation systems underperform in occluded areas due to inaccurate obstacle prediction and the computationally intensive process of creating ESDF maps. (b) Our HE-Nav system can generate energy-saving, collision-free and ESDF-free aerial-ground paths in real-time with the help of the LBSCNet model and AG-Planner.
  • Figure 2: HE-Nav system architecture. The perception module and path planner run asynchronously on the onboard computer, connected through a query-based map update method jmwang to ensure real-time local map updates with prediction results.
  • Figure 3: The overview of the proposed LBSCNet. It consists of semantic, completion and BEV fusion branches.
  • Figure 4: SCB-Fusion Module realizes the fusion of semantic features, geometric features and BEV features.
  • Figure 5: Illustration of AG-Planner and topological trajectory generation.
  • ...and 6 more figures