A novel navigation assistant method for substation inspection robot based on multisensory information fusion

Area of Science:

• Robotics
• Artificial Intelligence
• Sensor Fusion

Background:

• Substation inspection is time-consuming and labor-intensive due to complex environments.
• Current intelligent inspection robots rely on expensive high-precision LiDAR for navigation.
• Existing methods lack obstacle type identification and road contour boundary detection, limiting efficiency.

Purpose of the Study:

• To develop a novel, cost-effective navigation assistant method for substation inspection robots.
• To overcome limitations of high-precision LiDAR in obstacle identification and road contour detection.
• To improve inspection performance and efficiency through multisensory information fusion.

Main Methods:

• Implemented asynchronous information matching for multisensor data integration.
• Utilized 2D laser radar combined with an inertial measurement unit (IMU) for 3D obstacle imaging.
• Employed ENet for color point cloud segmentation using optical sensor data for environment perception.
• Developed the system on the Robot Operating System (ROS) and deployed it on an embedded platform.

Main Results:

• Reduced data volume by 95% compared to VLP-32C 3D laser radar sensors.
• Achieved 86% mean average accuracy in environment perception after training the ENet network.
• The navigation assistance module meets practical engineering application requirements.

Conclusions:

• The developed navigation assistance module enables substation inspection robots to identify road contours and obstacle types.
• The system successfully reduces data volume and hardware costs.
• Field tests in multiple substations confirm the robot's enhanced operational capabilities.