Design and Experimental Evaluation of EKF-LQR Navigation on Raspberry Pi for Autonomous Mobile Robots

Abstract

Accurate and robust navigation is essential for autonomous mobile robots operating in dynamic, GPS-degraded environments. We propose and validate a low-cost navigation framework implemented on Raspberry Pi, combining an Extended Kalman Filter for sensor fusion with a Linear Quadratic Regulator for optimal control. Unlike previous implementations, our system achieves real-time performance on resource-constrained hardware through computational optimization, maintaining localization accuracy during extended GPS outages of up to 30 seconds. The EKF fuses measurements from GPS, IMU, optical flow, and wheel encoders to provide reliable localization, while the LQR minimizes trajectory tracking error and control effort. Field experiments demonstrate that the proposed EKF–LQR approach achieves a Mean Square Error of 2.9 m², outperforming GPS-only and dead reckoning methods by 66% and 81%, respectively.