Intelligent Control With EKF-Based Multisensor Fusion for Differential-Drive Robots Subject to Unmodeled Dynamics

Abstract

Reliable control and state estimation of differential-drive robots (DDRs) operating in dynamic and uncertain environments remains a challenge, particularly when system dynamics are partially unknown and sensor measurements are prone to degradation. This study introduces a unified closed-loop estimation-control framework that integrates EKF-based multisensor fusion with Lyapunov-based RBF adaptive control. In this architecture, multisensor fusion improves the reliability of the robot state estimate under sensor noise, drift, and partial sensing degradation, whereas the radial basis function (RBF) adaptive controller compensates online for unmodeled dynamics and floor-induced disturbances. The proposed controller leverages the universal approximation property of neural networks to model unknown nonlinearities in real-time. An online adaptation scheme updates the weights of the RBF and the learned dynamics are integrated into a feedback linearization (FBL) control law. The Lyapunov analysis establishes closed-loop stability under standard boundedness and approximation assumptions, whereas the experiments demonstrate bounded tracking errors and improved performance under practical sensing and dynamic uncertainties. To ensure reliable state estimation, the EKF combines odometry from wheel encoders, inertial measurement unit (IMU), monocular vision, and 2D-LiDAR. The fused state estimate drives the intelligent controller, ensuring consistent performance even under drift, wheel slip, sensor noise, and failure. Gazebo simulations and hardware experiments on a differential-drive robotic platform demonstrated the effectiveness of the approach in terms of improved velocity tracking performance with a reduction in linear and angular velocity errors compared to the baseline FBL.