Multi-modal sensor fusion for highly accurate vehicle motion state estimation. (July 2020)

• Record Type: Journal Article
• Authors: Rodrigo Marco, Vicent; Kalkkuhl, Jens; Raisch, Jörg; Scholte, Wouter J.; Nijmeijer, Henk; Seel, Thomas
• Is Part Of: Control engineering practice. Volume 100(2020)
• Publication Date: 2020-07
• Subjects: Motion estimation -- Observability -- Automotive industry -- Non-linear systems -- Inertial sensors -- Kalman filter -- Odometry -- Collision avoidance -- Autonomous driving -- Simultaneous state and parameter estimation -- Systems and control engineering; Automatic control -- Periodicals; 629.89
• DOI: 10.1016/j.conengprac.2020.104409
• Languages: English
• ISSNs: 0967-0661

Abstract

Abstract: In the context of autonomous driving in urban environments accurate and reliable information about the vehicle motion is crucial. This article presents a multi-modal sensor fusion scheme that, based on standard production car sensors and an inertial measurement unit, estimates the three-dimensional vehicle velocity and attitude angles (pitch and roll). Moreover, in order to enhance the estimation accuracy, the scheme simultaneously estimates the gyroscope and accelerometer biases. The approach relies on a state-affine representation of a kinematic model with an additional measurement equation based on a single-track model. The sensor fusion scheme is built upon a recently proposed adaptive estimator, which allows a direct consideration of model uncertainties and sensor noise. In order to provide accurate estimates during collision avoidance manoeuvres, a measurement covariance adaptation is introduced, which reduces the influence of the single-track model when its information is superfluous. A validation using experimental data demonstrates the effectiveness of the method during both regular urban drives and collision avoidance manoeuvres. Highlights: Sensor fusion with 6-D IMU and standard production car sensors. Immersion transformation used to bring the system in a state affine form. Combination of both kinematic and dynamic models for enhance accuracy. Modified adaptive kalman filter used for vehicle motion estimation. Validation using real experimental data.