Hierarchical Control Architecture for Trajectory Tracking of Differential-Drive Nonholonomic Robots Using LQG-Based Velocity Control

Abstract

This paper proposes a hierarchical control architecture for trajectory tracking of differential-drive nonholonomic mobile robots intended for the Very Small Size Soccer (VSSS) competition. The proposed framework integrates a high-level nonlinear trajectory-tracking controller, which computes the desired linear and angular velocity references based on the robot pose and prescribed trajectory, with a low-level Linear Quadratic Gaussian (LQG) controller responsible for the independent velocity regulation of the drive motors. The LQG controller combines optimal state feedback through a Linear Quadratic Regulator (LQR) with integral action and state estimation provided by a Kalman filter, enabling accurate velocity regulation in the presence of modeling uncertainties and measurement noise. The complete architecture is implemented on an embedded differential-drive robotic platform equipped with incremental encoders, vision-based localization, and an ESP32-based controller, and is validated exclusively through real-time experimental tests under practical operating conditions. The experimental evaluation includes wheel velocity step-response tests and circular and lemniscate trajectory tracking under both forward and reverse motion. The results demonstrate accurate wheel velocity regulation, effective attenuation of encoder measurement noise, consistent tracking of the prescribed trajectories, and experimentally measurable trade-offs between spatial tracking accuracy, wheel velocity regulation, and control effort under different trajectory geometries and motion directions. Overall, the proposed architecture provides a modular hierarchical framework that integrates nonlinear high-level trajectory tracking with optimal low-level velocity regulation for real-time autonomous mobile robot applications.