In this paper a controller for anti-jerk control based on Neuro-Fuzzy models was introduced. First, the basics of Neuro-Fuzzy models and their structure have been introduced. The advantage of the Neuro-Fuzzy models is the possibility to build non linear structures in an easy way. Previous knowledge helps to define the structure, otherwise the algorithm itself is creating the different sub models according to the non linearities. The sub models themself are local linear models. With help of the Neuro-Fuzzy models, a drivetrain model to calculate the difference between engine speed and wheel speed was created. The model has sub models for tip in and tip out as well as for different engine speeds. For every sub model a controller with help of the root locus was designed. To simplify the application process an analogy to the classic controllers was introduced. So, the tip in requires a PD-controller, the tip out a simpler P-controller. Finally, tests in a vehicle have been made. The oscillations of the car can be reduced with help of the new controller for all engine speeds and for tip in and tip out. Subjective the anti-jerk controller with calculated speed difference was the better one. The next step will be a combination of the two approaches to get a closed loop with an observer for the difference between engine and wheel speed. Tests have to be made for different loads and driving situations as well as for other kinds of drivetrains, e.g. rear- wheel drive More improvement may be possible, if the controller will be calculated inside the engine control unit. because time lags for communication with the RPU can be excluded.
A neuro fuzzy approach for anti-jerk control
Ein Neuro-Fuzzy Näherungsverfahren für Anti-Ruck Regelung
2003
9 Seiten, 14 Bilder, 9 Quellen
Conference paper
English
A neuro fuzzy approach for anti-jerk control
Automotive engineering | 2003
|A Neuro Fuzzy Approach for Anti-Jerk Control
SAE Technical Papers | 2003
|2003-01-0361 A Neuro Fuzzy Approach for Anti-Jerk Control
British Library Conference Proceedings | 2003
|