Comparison of Two System Identification Approaches for a Four-Wheel Differential Robot Based on Velocity Command...

Area of Science:

• Robotics
• Control Systems
• Autonomous Navigation

Background:

• Precise modeling of differential drive robots is essential for autonomous systems.
• Two distinct modeling approaches, Motor-Based Model (MBM) and Simplified Model (SM), are evaluated for four-wheel differential drive robots.

Purpose of the Study:

• To comparatively analyze the accuracy and computational performance of MBM and SM for differential drive robot modeling.
• To assess the impact of model selection on odometry accuracy and navigation task design.

Main Methods:

• MBM utilizes four motor-specific transfer functions, while SM employs two transfer functions for linear and angular velocity.
• Both models were validated against real odometry data from a SLAM system on a differential-drive robot.

Main Results:

• MBM demonstrated higher accuracy in position and orientation estimation, with lower RMSE values compared to SM.
• SM exhibited superior computational performance, achieving a 30% reduction in simulation time and lower memory usage.
• MBM showed average position RMSE of 0.309 m vs. 0.414 m for SM; angular RMSE of 0.170 rad vs. 0.239 rad for SM.

Conclusions:

• MBM is more suitable for applications demanding high precision, such as SLAM.
• SM is advantageous for resource-constrained embedded systems requiring faster computation.
• The trade-off between model complexity, accuracy, and computational cost is critical for selecting appropriate robot models.