HE-Nav: A High-Performance and Efficient Navigation System for Aerial-Ground Robots in Cluttered Environments (2410.05079v1)

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/.

• The paper introduces HE-Nav, a system that uses a lightweight semantic scene completion network (LBSCNet) with attention and fusion techniques to enable real-time obstacle mapping.
• The paper's novel AG-Planner employs a kinodynamic A* search algorithm to achieve energy-efficient, safe trajectory planning without relying on computationally expensive ESDF maps.
• The experimental results report up to sevenfold energy savings and a 98% planning success rate, demonstrating significant performance improvements in cluttered environments.

The paper "HE-Nav: A High-Performance and Efficient Navigation System for Aerial-Ground Robots in Cluttered Environments" addresses the challenges faced by Aerial-Ground Robots (AGRs) in navigating through highly occluded and cluttered areas such as dense forests or environments with tall walls. Traditional systems using 3D semantic scene completion networks and Euclidean Signed Distance Field (ESDF) maps often struggle in these scenarios due to limitations in perception accuracy and high computational demands.

Key Contributions:

1. Perception Module:
   • The system introduces a lightweight semantic scene completion network (LBSCNet), which integrates a bird's eye view (BEV) feature fusion. This is further enhanced by a Sparse Contour Box Fusion (SCB-Fusion) module and an attention mechanism.
   • This module is designed to perform real-time and efficient obstacle prediction, creating a comprehensive local map of the environment even in areas with dense obstructions.
2. Path Planning:
   • HE-Nav incorporates a novel AG-Planner with a kinodynamic A* search algorithm, emphasizing energy efficiency. This planner ensures energy-saving path planning strategies while maintaining accurate trajectory predictions.
   • The trajectory optimization ensures paths that are safe, smooth, dynamically feasible, and avoids the use of ESDFs, improving overall computational efficiency.
3. Performance and Efficiency:
   • Extensive experiments have demonstrated that HE-Nav achieves up to seven times the energy savings in practical scenarios without compromising planning success, which stands at a remarkable 98% in simulations.
   • This system's ability to process and navigate through cluttered environments effectively highlights significant advancements in both perception and path planning for AGRs.

The research presents a comprehensive solution that improves both the efficiency and performance of AGRs in challenging environments, making significant strides in the robotics field. The code and additional resources are made available by the authors for further exploration and use in related applications on their project page.