Decentralized Fusion-Based Ego Velocity Estimation Using Multiple FMCW Radars

Abstract

This article presents a decentralized fusion-based ego velocity estimation framework utilizing multiple Frequency-Modulated Continuous-Wave (FMCW) radar sensors and an Inertial Measurement Unit (IMU). This method relies on the collaborative exchange of information among radar sensors, employing a decentralized estimation approach. This approach enhances system robustness and incorporates a unique and effective noise reduction technique to diagnose and eliminate undesirable measurements. Traditional robotics missions, relying on visual data, often face challenges, particularly in scenarios where Global Navigation Satellite System (GNSS) signals are either not accessible or unreliable and when visibility is hindered by factors such as darkness, direct sunlight, fog, or smoke. The proposed framework holds immense potential for enhancing the precision of navigation systems in autonomous vehicles and advanced driver assistance systems, particularly in challenging conditions like adverse weather and complex urban environments. Comparative analysis against traditional centralized fusion methods in experimental studies reveals sub-stantial enhancements in accuracy and reliability, highlighting the potential of decentralized fusion.