A Low-Cost Multi-Purpose Localization Design for Land Vehicles

This paper presents a novel, low-cost, multi-purpose localization algorithm designed for land vehicles, capable of estimating relevant states across three distinct domains: vehicle dynamics and control, navigation, and advanced driver assistance systems (ADAS). Traditionally, these domains rely on separate estimators despite having overlapping states of interest, but our algorithm unifies the estimation process to provide all necessary states from one source. The proposed algorithm achieves low cost by reducing computational demands and utilizing a standard sensor suite commonly found in modern vehicles, including wheel speed sensors, steering angle sensors, a 6-degree-of-freedom (DOF) inertial measurement unit (IMU), and GNSS position and velocity. To address the unique challenges of each domain, we introduce several specialized techniques: a neural network-based wheel slip detection algorithm and a “no-wheel-left-behind” strategy for optimal wheel speed selection for vehicle control use cases, a GNSS rejection and recovery mechanism employing the normalized innovation square metric and shape matching technique for navigation use cases, and a comprehensive state-machine design to maximize output availability for all applications. Our algorithm has been successfully deployed across all Rivian electric vehicles, demonstrating its versatility and effectiveness in diverse real-world scenarios.