An Automated Parameter Search Method for Vehicle Transmission System

Abstract

In the vehicle industry, shift quality is a major factor to evaluate the performance of a transmission unit. Following by the electric control units like ECU (Engine Control Unit) and TCU (Transmission Control Unit) took important role in the vehicle transmission, the computational parameters in the control units also took the key role to determine a shift quality. To find a set of optimal parameters, trial-and-error technique is most famous in the field. However the technique requires human experience for efficient and qualified parameter search. To generalize the quality and search time, there is a certain demand on the automation of parameter search. The studies for automated calibration of automatic transmission system for shift quality have been placed in the automobile field. Actual vehicle and Dynamometer are used to automate the system, and they are based on the technique called Design of Experiment. Since the vehicle transmission systems are complex, exhaustive search technique is not proper to solve the problem. Design of Experiment limited parameters to handle only key factors. However the technique still requires human experience at the design level. In this thesis, we propose a genetic algorithm based transmission parameter search method for a vehicle transmission system. In the pre-process phase, learn each parameters effect on the target shift quality to avoid the dependency on human experience. Global search with genetic algorithm approach and local optimization is implemented. Simulink based vehicle simulator is used for experiment.