Toward a Low-Cost Planetary Rover: Real-Time Embedded Multi-Sensor Fusion for Robust Obstacle-Aware Navigation

This paper presents the design, real-time implementation, and experimental validation of a low-cost, embedded multi-sensor fusion framework for dead reckoning, localization, and obstacle detection in mobile ground robots. The proposed system integrates a MEMS inertial measurement unit (IMU), a GPS receiver, and an ultrasonic sensor (HC–SR04) within a single synchronized execution loop to achieve GPS-resilient and obstacle-aware navigation on resource-constrained hardware. A hybrid complementary–Kalman filtering scheme is implemented, where the complementary filter fuses accelerometer and gyroscope data for roll and pitch estimation, while periodic GPS updates perform lightweight Kalman corrections to bound drift. Ultrasonic range measurements are processed in the same loop and mapped relative to the estimated trajectory to provide continuous hazard awareness. The complete system, implemented on a 16 MHz Arduino-class MCU, achieves real-time operation at 20 Hz with a total hardware cost below USD 150. Field trials conducted under mixed GPS conditions demonstrate that the fused solution outperforms individual sensors, achieving a position RMSE of 1.19 m—a 49.6% improvement over GPS-only and an 80.7% improvement over IMU-only estimates—and maintains 99.2% localization continuity during GPS outages. The proposed framework demonstrates that reliable, low-cost, and real-time multi-sensor navigation can be achieved without high-performance processors, offering a practical architecture for educational robotics, cost-sensitive autonomous vehicles, and preliminary planetary rover prototypes.