Localization of solar panel cleaning robot combining vision processing and extended Kalman filter

Area of Science:

• Robotics
• Computer Vision
• Sensor Fusion

Background:

• Accurate localization is critical for autonomous mobile robots, especially in structured environments like solar farms.
• Existing methods for solar panel cleaning robots often struggle with precision, particularly at higher operating speeds.

Purpose of the Study:

• To develop and validate an improved position estimation method for self-driving solar panel-cleaning mobile robots.
• To enhance localization accuracy beyond traditional vision-based line counting.

Main Methods:

• Utilized image processing for line counting on the solar panel floor, with adjusted thresholds and speed-dependent offsets.
• Integrated inertial measurement unit (IMU) data and wheel encoder information for enhanced position estimation.
• Employed an extended Kalman filter, fusing IMU and encoder data for precise localization, especially between panel lines.

Main Results:

• Achieved a Root Mean Square Error (RMSE) accuracy of up to 0.089 m at a speed of 100 mm/s.
• The integrated method demonstrated nearly 50% improvement in accuracy compared to vision-based line counting alone.
• IMU signals effectively differentiated horizontal and vertical lines based on robot heading and detected line angles.

Conclusions:

• The proposed position estimation method significantly enhances the accuracy and reliability of solar panel cleaning robots.
• The fusion of vision, encoder, and IMU data via an extended Kalman filter offers a robust solution for autonomous navigation in solar farms.
• This advancement contributes to more efficient and precise robotic operations in renewable energy infrastructure.