Worakit Puangsup. Vehicle parameter estimation using recursive least square and extended state observer. Master's Degree(Mechanical Engineering). Mahidol University. Mahidol University Library and Knowledge Center. : Mahidol University, 2019.
Vehicle parameter estimation using recursive least square and extended state observer
Abstract:
The model-based control has been widely used to improve vehicle operational performance. The difficulties of applying the model-based approach include not only to find a suitable model, but also to obtain numerical values of all related parameters. Moreover, in most cases, the numerical value of parameters can be changed in real life. For some fortunate cases, measuring the change can be done but not in real time, or not without additional cost. The worst case is when the change is not known or unmeasurable. The mismatch between the model and the real life can degrade the performance of the designed model-based control. General model-based automotive transmission control for longitudinal vehicle motion usually relies on a single-degree-of-freedom longitudinal vehicle model. This model consists of the driving force from the engine to moving the vehicle, which is considered as a lumped mass, and subjected to resistant forces include the aerodynamic drag force and the resistance due to the road surface and slope. Unfortunately, the vehicle mass can change depending on the number of passenger and the carrying load. The air density or temperature also changes which affects the aerodynamic force. And of course, the road surface and slope are also variable. The measurement of these variations cannot be done in real time or onboard due to the unavailability of the suitable sensor and the additional cost. Hence, the performance of the designed model-based control may not be at its best as the value of parameters is incorrect. This research focused on the development of an estimation method that can estimate the vehicle parameters describing the longitudinal vehicle dynamic motion. To be specific, this research focused on identifying the vehicle mass, the aerodynamic drag force, and the road slope simultaneously in real time. To do so, the Recursive Least Square (RLS) with forgetting technique is applied. The accurate knowledge of the aforementioned parameters also leads to the ability to estimate vehicle load torque, which is very beneficial especially for transmission shift quality control. The adaptive Extended State Observer (ESO) is applied to this part, which relies heavily on the results from RLS. The proposed design was studied through both simulation and experimentation. The proposed scheme performed well both in simulation and experimentation. Due to limited resources, only the estimation results for the vehicle mass and road slope were validated experimentally. Thus, the estimations of the aerodynamic drag force and the vehicle load torque showed reasonable responses and numerical values.