How can we address the issues of wheel slippage and loss of positioning in robot chassis?
Smooth epoxy floors and oil stains are the root causes of slipping and loss of positioning for AGVs and AMRs. In settings such as warehousing and logistics, 3C electronics, food processing, and cleanrooms, the chassis of mobile robots frequently experience slipping and loss of positioning.
Mobile robot chassis typically employ fixed or shock-absorbing designs, which perform exceptionally well on dry concrete floors but frequently malfunction in the following scenarios:
Machine shops with oil spills and coolant splashes: The wheels have low friction with the floor, making them highly prone to skidding during acceleration or turns;
Slippery floors in food and beverage plants: Residual water films or grease left after cleaning cause drive wheels to spin freely;
Flammable and explosive areas in chemical and pharmaceutical facilities: Slipping not only reduces efficiency but may also lead to collisions or safety incidents;
Metal or epoxy floors with steep inclines or severe wear: Positioning accuracy derived from the fusion of traditional odometers and IMUs drifts rapidly, dropping from centimeter-level to meter-level precision.
Domino Effect of Position Loss
Odometer Drift: The chassis relies on wheel encoders to calculate its position. Slippage causes a discrepancy between the encoder data and the actual trajectory; this error accumulates over time, ultimately causing misalignment in SLAM mapping.
Sensor Failure: Highly reflective surfaces scatter LiDAR beams, while visual cameras fail to match due to a lack of textural features. Although the IMU can provide short-term compensation, it cannot eliminate long-distance drift.
Typical Symptoms: Path deviation, repeated re-localization, emergency stop alarms, and even collisions with other devices.
Active Intervention Strategy in Anti-Slip Mode
By integrating wheel speed encoders, an IMU, vision/laser SLAM, and a ground friction estimation model, the algorithm can detect slippage in real time. Once a mismatch is detected between wheel speed and actual movement speed, the system immediately switches to anti-slip mode:
Dynamically adjusts motor torque output to prevent wheel spin;
Based on real-time friction estimation results, the system employs torque vectoring control combined with PID adaptive tuning to rapidly reduce drive wheel torque output, preventing high-speed wheel spin. Simultaneously, it performs differential fine-tuning on the left and right drive wheels to maintain the robot’s posture stability.
Visual odometry and multi-feature point matching are employed to correct positioning drift;
A depth camera captures rich feature points such as ground textures, edges of oil stains, and metal scratches to perform optical flow tracking and matching using multiple feature points.
Even in areas covered by oil stains, the system can achieve centimeter-level positioning correction by extracting high-contrast edges and texture-invariant features.
By using machine learning models to predict the coefficient of friction for different surface materials, the system enables early warning and optimized route planning.
The algorithm incorporates a lightweight machine learning model that continuously learns the friction characteristics of various surface materials (such as concrete, epoxy resin, metal, oil films, and water films) to generate a real-time friction coefficient map.
