Real Time Vehicle State Estimation using Error Map Function

The estimation of vehicle handling and control parameters in dynamic conditions is challenging due to errors and delays in real-time data logging with low-resolution onboard sensors. These issues significantly impact the performance of vehicle stability and control algorithms, particularly in vehicles under testing. This study presents a statistical approach for real-time vehicle state estimation that addresses the limitations of low-resolution sensors with errors and delays in measured signal. In this study, a real-time (RT) model was developed and trained using an in-house electric SUV to estimate yaw velocity and slip angle. The model leverages other measured signals available from the vehicle’s onboard sensor setup. It integrates an error and delay function with an error predictive model to estimate the targeted parameters' signal response in real time. The RT model introduces an error function method that enhances prediction accuracy by combining the error map value with the error weight of the target signal. The predictive model uses multi-linear regression based on high-resolution sensor data to improve the estimation of signals that are either not logged or inaccurately logged in low-resolution setups. Model training and error map assignment is an iterative process that utilizes on-track vehicle test logs from two data measurement sources. In each iteration, the error map is updated, assigning error weights to the parameters, making the model adaptive to error characteristics during subsequent training. The model was validated through ISO ramp steer and ISO chirp tests, demonstrating its effectiveness in enhancing real-time vehicle state estimation using error map function.