Towards a Generic Architecture for Multi-Vehicle Autonomy

We describe the motivation for a programme of research “MP012: Software Architecture for Hybrid Decision Making” that is being undertaken in SEAS DTC Year 2. There are many ways to endow an autonomous system with intelligent behaviour, including planning, machine learning, and rule-based methods. Each of these has advantages in certain settings (scenarios, mission phases, etc) that we will aim to characterise. Therefore a hybrid architecture that can closely integrate multiple techniques is required. In order to achieve the integration effectively, the software architecture must be based on a formal understanding of sequential decision-making (optimal control). We also discuss how synthetic environment demonstration of the proposed architecture may be facilitated under the TIED theme. © QinetiQ, 2006.