Architectures for intelligent vehicles

Abstract

Architectures decompose complex engineering problems into smaller, more manageable pieces, the modules of the architecture. The objective is to engineer, implement, and test the modules separately. If the modules have well-defined interfaces and implement well-defined services, it is possible to modify some of the modules, without modifying them all. As intelligent vehicles evolve, because of better sensors, better control mechanisms, and new applications, most of the modules in the previous implementation can be reused, rather than requiring an entirely new implementation. The modules in the architecture also isolate operations that require specialized knowledge. In intelligent vehicles, some operations require knowledge of the physics of motion of the vehicle, the stability of different control mechanisms, or the limitations of different sensing mechanisms. By performing operations that require specialized knowledge in independent modules, experts on a topic can implement the appropriate modules without requiring any one person to be expert in everything. The architectures that are being designed for intelligent vehicles are changing from simple interconnections of modules to layered architectures, similar to those used in communications networks. A layered architecture has successfully allowed the Internet to evolve over the past 50 years. Intelligent vehicles are more complex than communications networks. They interact with the physical world in many ways, and collaborative maneuvers are time critical. An architecture that satisfies the new requirements, while preserving the characteristics of traditional architectures, is described.