Vehicle side-slip angle estimation with deep neural network and sensor data fusion

Modern chassis control systems, advanced driver assistance systems (ADAS) and automated driving systems that demand a precise vehicle localization or a reasonable trajectory planning desire a highly accurate and reliable vehicle state estimation. However, the traditional methods such as Kalman and RLS filter, which based on the vehicle dynamic model, mainly rely on the differential equations that approximate the vehicle behaviour in reality [1, 27, 31]. The vehicle dynamics is such a nonlinear and multidimensional system with numerous parameters, which makes it very difficult to adapt the parameters in different situations and figure out appropriate model equations.