Real-Time Collision Mitigation Strategies for Autonomous Vehicles

Autonomous vehicles (AVs) offer a promising solution to mitigate human errors and, consequently, reduce the number of road traffic fatalities. While AVs may significantly reduce the number of collisions in the future, they will not be able to completely eliminate it, as certain accidents cannot be avoided due to the physical limitations of the vehicle, and the boundary conditions imposed by the environment. This paper presents an initial investigation into a novel methodology that aims to mitigate the consequences of unavoidable accidents by determining the vehicle trajectory that minimizes injuries to the occupants of the AV in real-time. This result is achieved using a multibody (MBD) crash simulation software to simulate all the possible impacts and determine the one with the lower severity. The approach used is hybrid: the AV can run simulations in real-time as well as access a database where pre-simulated crash results are stored. The maximum computational time required by the software to perform an impact simulation was 80 milliseconds, comfortably within the total time allowable for the AV to detect the unavoidable accident and implement corrective actions, determined to be 0.3 seconds for the scenarios considered. The results of the multibody code were compared with a finite element code, and among the severity indices considered, the Theoretical Head Impact Velocity index has the highest correlation coefficient between the two software. Overall, the proposed methodology demonstrated potentiality both in terms of results and feasibility and could have significant implications for the development of safer self-driving cars.