Slip-Correction Methods for Improved Dead Reckoning in Autonomous Wheeled Robots
Loading...
Authors
Diab, Jude
Date
2025-05-02
Type
thesis
Language
Keyword
Slip Correction , Current-based slip correction
Alternative Title
Abstract
This work compares methods for improving the dead-reckoning capabilities of a mobile robot in the presence of longitudinal slip. They were evaluated using a standardized test of driving in a 140-centimetre square. Testing was conducted on high-friction rubber and low-friction artificial ice surfaces, with a motion-capture system providing ground-truth data for evaluation. Results were both assessed visually, and using RMSE of the motion-captured path against the desired path. A custom-designed robot served as a standardized hardware platform for testing.
Five methods were studied. First, an uncorrected method, measuring only odometry and incorporating a lateral-position correction and wheel-velocity control, showed poor path-following performance over a 140 cm square path, with RMSE values of 48.5 cm on rubber and 45.8 cm on ice, despite tuning. The results of this method were used as the benchmark. Second, a current-slip method, correlated motor current to velocity loss due to slip, achieved RMSE values of 22.2 cm on rubber and 30.9 cm on ice. Third, a torque-slip method, which incorporated torque measurements instead, showed little improvement on the uncorrected method, with RMSE values of 43.2 cm on rubber and 53.5 cm on ice. Fourth, by using a current-slip-ice method that correlates data that is specific to ice, RMSE values as low as 11.9 cm were achieved. However, the exact corrected path showed variability, with a coefficient of variation of 25% across three runs. Fifth, the equivalent test carried out for torque, the torque-slip-ice method yielded almost no improvement, with an RMSE value on ice of 52.7 cm.
In summary, the current-slip method showed the best improvement in performance under the conditions tested. These findings highlight the potential of a current-slip method for mobile robotic navigation in low-friction environments, with implications for future research in mixed and uneven terrain.