Hitch Force Estimation for Electric Caravan with Deep Learning Method

The determination of the hitch force at the articulation point in the vehicle-caravan system is an important parameter that affects the stability of these systems. Especially in the case of electric propulsion generation in the caravan system, determining the effect of this electric propulsion on the vehicle emerges as a data that must be obtained directly or indirectly. In this paper, a deep neural network (DNN) is designed for hitch force estimation. It is modelled to better understand the forces acting on the vehicle-caravan system. The inputs to be applied to the DNN have been selected to consist of parameters affecting the hitch force. While estimating the hitch force at the articulation point, only the sensors in the caravan are used. According to the field test results consisting of 6.09 km, it has been shown that with a DNN, the hitch force can be predicted with an error of 12.26% using only the sensors in the caravan. Compared to existing model-based techniques that achieve an error of 9.5% using inertial measurement unit (IMU) and global positioning system (GPS) sensors in the towing vehicle, the proposed method is considered a practical and sensor-efficient option. The obtained results confirm that DNN-based prediction methods can be an alternative technique for vehicle-caravan systems and show the potential for further accuracy improvements through additional training data and different test scenarios.