Adaptive error and sensor management for autonomous vehicles: Model-based approach and run-time system

Over the past few years semi-autonomous driving functionality was introduced in the automotive market, and this trend continues towards fully autonomous cars. While in autonomous vehicles data from various types of sensors realize the new highly safety critical autonomous functionality, the already complex system architecture faces the challenge of designing highly reliable and safe autonomous driving system. Since sensors are prone to intermittent faults, using different sensors is better and more cost effective than duplicating the same sensor type because of diversity of reaction of different sensor typesto the same environmental condition. Specifying and validating sensors and providing technical means that enable usage of data from different sensors in case of failures is a challenging, time-consuming and error-prone task for engineers. Therefore, in this paper we present our model-based approach and a runtime system that improves the safety of autonomous driving systems by providing reusable framework managing different sensor setups in a vehicle in a case of a error. Moreover, the solution that we provide enables adaptive graceful degradation and reconfiguration by effective use of the system resources. At the end we explain in an example when and how the approach can be applied.