Debugging the Networked Car

Abstract

In the automotive industry, modern cars are equipped with numerous sensors and actuators to control safety critical aspects. These sensors collect large amounts of information, which are processed and used by vehicles to control brakes, steering, engine, etc. The AU-TOSAR standard specifies the implementation details of these software units. This approach provides more flexibility for the management of vehicular Electrical/Electronic(E/E) systems related to complex driver assistant functions. However, due to the distributed features of such systems, software programs built on the AUTOSAR platform are difficult to achieve high quality and correctness. AUTOSAR programs are running on different machines and exchange messages via the underlying in-vehicle network, which makes the system non-deterministic and non-predictable. In this thesis work, these problems are addressed and a new debugging tool based on a symbolic execution engine, named KLEE, is developed. This new tool enables effective testing of AUTOSAR programs before deployment on Electronic control units(ECUs). It has the capability to detect pointer-related problems and inject non-deterministic failures to programs while symbolically executing them. We explore the design and the implementation of this tool and also come up with several possible scenarios as use cases of this debugging tool.