Modular multi sensor data fusion architecture for autonomous vehicles: Using the RACE system on a segway RMP 440

Recent developments in centralized ICT architectures have realized the abstraction of hardware and communications functionality into software within for of highly assisted and autonomous vehicles. This development requires a data-centric approach to the integration and fusion of sensory information within the vehicle. We address this requirement through utilizing a hierarchical framework of intelligent agents. While each agent maintains a well-defined inputs and outputs, the internal processing can be modified to influence the behavior of the system and fulfil the requirements of the application. The hierarchical structure of these agents allows the system to transition from a data-dense environment at the sensor layer to information-dense environment at the application layer. This enables the integration of disparate sensors, system models and sensor fusion in a modular and non-monolithic fashion which is no longer intimately linked to the application.