Development and Validation for Extended Vehicle State and Load Estimation

Vehicle state and vehicle load estimation play a critical role in vehicle testing and advanced driver assistance systems, both presently and in the future. However, existing methodologies tend to perform poorly when faced with common external disturbances, such as varying road angles, sensor noise and bias, varying tire-road friction, wheel slip, and even unknown vehicle mass.

In this study, a series of extended vehicle state and load estimators are investigated to address these challenges and provide more reliable estimations. To achieve this, increasingly complex models are considered for use in the estimators, allowing for increasingly more corrections of external disturbances and enabling sensor fusion across a wider range of sensors.

The research begins with a review of the general estimation framework, also referred to as the virtual-sensor framework. Subsequently, estimators based on a planar kinematic model, a planar dynamic model, a model incorporating wheel rotations, and finally, a model incorporating three-dimensional motion are explored.

The developed estimators are then compared and experimentally validated using measurement data from a Range Rover Evoque. Key findings indicate that as model complexity increases, the estimators become robust against a wider range of disturbances, including varying road angles, sensor noise and bias, varying tire-road friction, wheel slip, and even unknown vehicle mass. However, it is important to note that increased model complexity introduces additional frequency modes and non-linearity, necessitating finer sampling. Furthermore, the increased number of parameters and states associated with higher model complexity requires careful attention when tuning. Nonetheless, this increased tuning set has proven useful in enhancing estimator performance, especially when employing heuristics, for example.