Innovation in Automotive concerns new functionalities in vehicles or even new ways to accomplish some old features reducing their costs.
Anti-pinch protection systems, for instance in window lifting or seat sliding, may require long activities in order to fit the software control logic for their mechatronic systems increasing development costs.
EMA proposes to replace traditional software solution with a machine learning (ML) approach, which automatically models the electro-mechanical environment. Although ML is already being used in some automotive areas, e.g. traction control units, EMA objective is to explore other application domain areas where ML can be exploited.
EMA developed an automated testbench that emulates an automotive electric seat, with different obstacles profiles. A ML model has been trained and implemented on a low-performance MCU. The result was comparable with the traditional development approach, with the advantage of minor developing time since no calibration is needed.
