Systematic Model-Based Design and Implementation of Supervisors for Advanced Driver Assistance Systems

Abstract

The number of advanced driver assistance systems (ADASs) and the level of automation in modern vehicles is increasing at a rapid pace. Moreover, multiple of these ADASs can be active at the same time and therefore may need to interact with each other. As a consequence, the design of the supervisor layer that is responsible for proper coordination of the control tasks performed by the low-level ADASs controllers is becoming more complex and safety-critical. For this reason, there is a strong need for automated synthesis tools that lead to supervisors that are safe by design. In this paper, we present a systematic approach to model-based supervisor design using discrete-event system representations. In particular, this paper shows that the proposed method is suitable to deal with the multiple and complex systems of interacting ADASs. To be more specific, in contrast to current practice, which often relies on textual specifications and exhaustive testing, the proposed method has four main advantages: 1) it is based on mathematically specified requirements that only allow one interpretation; 2) it prevents blocking situations by design; 3) it guarantees correctness in the sense that the resulting supervisor satisfies all the specified requirements; and 4) code is generated from the obtained supervisor which eliminates the need for manual coding. The proposed method is demonstrated by means of a case study on cruise control and adaptive cruise control. The resulting supervisor is validated by simulations and experiments on a modern passenger vehicle. Based on the results presented in this paper, it can be concluded that the model-based supervisor design, simulation, and implementation method is promising and powerful for future applications in the automated vehicle systems.