Sensor configurations and testbed for vehicle state estimation

Abstract

This paper conducts an investigation on whether vehicle dynamic states such as pitch, roll, yaw, and particularly sideslip, can be estimated using limited and inexpensive automotive-grade sensors without the use of model based estimators (parameter free). First, multiple sensors and combinations of these sensors are evaluated for the purpose of estimating vehicle states with different levels of accuracy. Second, experimental tests are carried out on a purpose built vehicle. As most commercial vehicles come equipped with an Accelerometer and Gyroscope (INS), we use this sensor combination along with a single antenna GPS receiver for our tests. The state estimation algorithms for the given sensor combination are implemented in a dSPACE MicroAutoBox that allows real-time computation. The test vehicle is a high-power radio-controlled car, which also carries a commercial data logger that is used to set a benchmark for the estimated states.