Blending Methodologies – Simulator and Prototypes @UniBO

The design of most subsystems of a vehicle are becoming more and more critical, because of the timing for the modern projects. The subsystem related to the active chassis controls belongs to that basket and their design and refinement is becoming even more critical with the new non-traditional powertrain architectures. In such applications, indeed, the integration of all the vehicle subsystems increases its complexity.
The presentation shows a new method to merge together the typical offline simulation with track data measurements and prevision of the expected performance of new components design done by a DiM150 dynamic driving simulator (see https://lnkd.in/d8uQyQza).
In particular, the presentation shows an overview of the Danisi Engineering’s blended methodology based on the virtual mules: simulation models suitable to be driven in real time on the driving simulators. The methodology enables a mirroring of the vehicle development approach: from the classic one (widely based on physical prototypes) to the new blended method based on the virtual mule, comprehensive of hardware in the loop and software in the loop applications.
The presentation reports some application examples, in which the described methodology is applied embedding one or more components based on Artificial Intelligence (AI) or Machine Learning (ML).
Applications range from obstacle avoidance with Reinforcement Learning, where the RL training of the agent is based on vehicle dynamics simulations, to the development of a traction control that uses an ML vehicle model, which is tuned and validated with other vehicular systems using driving simulators.