Combining Behavior-Based and Contract-Based Control Architectures for Behavior Optimization of Networked Autonomous Vehicles in Unstructured Environments

Abstract

The driving behavior of networked autonomous vehicles can be optimized with collaborative knowledge by sharing safety-relevant runtime data and reliably integrating it into the individual vehicle platforms. Since the highly dynamic application context in the vehicle domain is safety-critical, the behavior adaptation process must be performed in a correct and predictable manner despite networking, reconfiguration and integration of external services. The Dynamic Safety Contracts (DSCs) approach provides a systematic, methodical and verifiable composition concept for the integration of external services at runtime in order to become certifiable for future applications. For a dedicated collaborative mode, a virtual safety cage can be specified with DSCs to derive the permitted degrees of freedom. In hazardous situations while driving, a predefined safe reaction behavior can be triggered in this way. Especially in highly unstructured environments like the off-road domain, collaborative knowledge such as additional spatial perspectives can improve the individual capabilities of environmental perception and ultimately the overall system behavior. For this challenging application domain a robust autonomous control like the integrated Behavior- Based Control (iB2C) is most suitable. In this paper, we show the benefits of a combined control architecture.