A virtual sideslip sensor has been developed for automotive applications, to improve active chassis control systems devoted to increase vehicle safety and performance. High costs and scarce reliability of conventional sideslip sensors do not allow their employment on passenger vehicles, while approximated estimation methods are effective only when the vehicle lateral dynamics is confined in the linear field. The sideslip angle is defined as the angle between the vehicle longitudinal axis and the velocity vector measured in the centre of gravity. In steady state manoeuvres, e.g. a ramp steer, the sideslip angle increases linearly with the lateral acceleration only in the linear vehicle operating range. This trend is not affected by the road surface condition, neither by the vehicle speed. For high lateral acceleration values, the relationships become strongly non linear, while the influence of some state variables increases. Moreover, the road surface condition becomes critical, affecting dramatically the lateral vehicle dynamics, through the tire-terrain friction coefficient. The friction factor is the external parameter which affects the dynamic lateral dynamics, reducing the lateral acceleration range while increasing the corresponding sideslip angle, according to a nonlinear relation. Moreover it cannot be measured on board. For this reason a cluster approach has been used, based on a manual switch which allows the driver to choose between dry&wet or snow&ice asphalt conditions. A neural virtual sideslip sensor has been developed for large scale employment on passenger vehicles. Results clearly shows the sensor feasibility as well as its reliability over a very wide range of driving conducts (normal/sport), load and road surface conditions. Different architectures have been tested, with an increasing number of input and output parameters. Precision is improved mostly by a massive training activity on medium/slow and high speed manoeuvres. Although this method is effective even when the lateral vehicle dynamics has a strongly non linear behaviour, it fails for non nominal pressure conditions, which imply radical dynamics alterations.
Non-adaptive neural automotive sideslip virtual sensor
Nichtadaptiver, virtueller Sensor für die Schätzung des Drehwinkels eines Straßenfahrzeuges unter Verwendung eines neuronalen Netzes
2004
6 Seiten, 6 Bilder, 5 Tabellen, 3 Quellen
Conference paper
English
Virtual sensor: application to vehicle sideslip angle and transversal forces
Tema Archive | 2004
|Automobile sideslip prevention device and sideslip prevention method
European Patent Office | 2023
|Online Contents | 1998
A Deep Learning based Virtual Sensor for Vehicle Sideslip Angle Estimation: Experimental Results
British Library Conference Proceedings | 2018
|