Adaptive Steering System for Improved User Experience

A car's steering is the first touchpoint for the driver and one of the most important factor which governs the buying decision of the vehicle. Currently steering system has a single EPAS map leading to constant steering feel during both low and high speed manoeuvres. Some vehicles are equipped with steering system having fixed driving modes which need manual intervention. In order to achieve the needs and to overcome the limitation of fixed mode steering system, this paper proposes Machine Learning based Adaptive Steering System which automatically switches from comfort to sports mode and vice a versa to provide soft and hard steering feel as per the user's requirement. The system identifies driving pattern and zone based on vehicle response to driver input and accordingly adapts steering assist / torque. The system comprises of two modules namely driving pattern recognition for sport and comfort mode and zone detection for city and highway driving. Both modules used already available vehicle CAN data for the system functionality. Decision tree algorithm is used for classifying driving pattern into sports and comfort mode with the accuracy of 98% and time series clustering is utilized to identify driving zone for mode holding. The system is successfully validated using double lane change and high speed cornering like manoeuvres.