Reliability analysis and optimization of ADAS systems

Alexandre Mugnai, Alberto Clarich (ESTECO), Diego Minen, Guido Bairati (VI-Grade)

CHALLENGE -  ADAS (Advanced Driving Assistance Systems) systems are integrated systems aimed to increase the passive and active safety performance of a car and these systems include sensors, signal processing and filtering, controllers and actuators at braking, suspension and engine level. The validation of ADAS systems today is still a difficult task to achieve as it is unclear what the load cases getting the system to fault are considering that the ADAS system needs to be operational all the time, not to mention the need to fulfil the Euro NCAP requirements. It is therefore mandatory to develop a simulations approach in which the ADAS control systems, integrated with sensors and actuators, vehicle dynamics, inter-vehicle behavior and traffic environment are simulated. 

SOLUTION -  VI-CarRealTime, SCANeR and modeFRONTIER software are integrated for the reliability analysis of an ADAS system, proving how fault tolerance can be identified and optimized for any combination of control system and traffic scenario.  The sensor simulations and 3D world environment are performed with SCANeR (Oktal) and combined with the VI-CarRealTime (Vi-Grade) vehicle dynamics simulations to evaluate the car behavior under the given conditions. The simulation models are then integrated in the modeFRONTIER process automation workflow, allowing the automatic execution of the simulation analysis under the various design parameters.

Once the process workflow is defined, simulations can be executed automatically by modeFRONTIER.  Reliability and Robust Design tool can be used to quantify accurately the distribution of a stochastic response,  in presence of input parameters uncertainties. The population is sampled by a Latin Hypercube DOE, accordingly to the distribution of the input parameters, and the distribution of the response is evaluated by the Polynomial Chaos Expansion (PCE), which allows to evaluate analytically mean, standard deviation and any percentile value of the response distribution.

BENEFITS - The work conducted here allowed the assessment of the ADAS function assessment emphasizing the scenarios in which performance improvements are needed by further tuning the sensing – algorithm – actuation aspects of the function.​