Estimating the Road Friction Coefficient Dynamically using Machine Learning

In autonomous vehicles, it is vital for the vehicle to drive in a manner that ensures the driver is comfortable and has confidence in the system, which ensures he does not feel compelled to intervene or take control of the vehicle. The system must consider environmental factors and other aspects to provide the driver with a comfortable and stress-free drive. In this regard, the road friction coefficient, which quantifies the grip experienced by the tire on a road, is a critical parameter to be considered by several comfort and safety functions. An inaccurate estimation of road friction coefficient can lead to discomfort in worst case safety risks for the driver, as the system would be over or underestimating the tire’s grip on the road and this alters the vehicle’s response to control inputs. In the context of Advanced Driver Assistance Systems (ADAS), dynamically estimating the road friction coefficient can significantly improve the safety and comfort of driving functions. However, estimation of the road friction coefficient dynamically requires complex mathematical modelling of nonlinear relationships that are challenging to solve by numerical or analytical methods. This, coupled with the need for real-time estimation, presents a noteworthy challenge in practical systems. As on date, state-of-the-art driving functions often assume a fixed friction value or operate within a friction range. Therefore, we propose a multivariate time-series model to dynamically estimate the road friction coefficient. The model is trained on a synthetic dataset comprising vehicle dynamics data for cars driven on roads with different road friction coefficients. We evaluate multiple model architectures and hyperparameter configurations against metrics like accuracy and inference time to identify the best-performing model. Furthermore, the models are trained with input signals that are available in the vehicle, so they are suitable to be deployed in real-world contexts.