Network-Aware Mission Planning for Reliable Autonomous Robotic Inspection Missions

Brief description of the context of the research including potential impact

Autonomous mobile robots for inspection, maintenance and repair (IMR) are capable of operating outside of the coverage of wireless networks, but rely on such networks to transmit data, receive new commands, and for failure recovery (e.g. via teleoperation). This reliance is typically implicit in a robot's control software, leading to errors when the network is not available.

In the worst case this could lead to a robot becoming stuck in a dangerous position but unable to communicate with an operator. To address this, we will develop a mission planning system which selects robot actions (e.g. navigation routes) using a model of network coverage to minimise the risk of failure, e.g. adopting conservative motion parameters when out of communication range.

We will then explore the planning algorithms which are able to position network access points as a robot moves, allowing it to guarantee network coverage for actions which have a higher likelihood of failure.

Industry will benefit from this research through the creation of more robust robot behaviour, allowing less monitoring of deployed systems and an increase in trust in their usage, particularly in areas with less structure or unreliable or uncertain network coverage. This will lead to an increase in number of problems where robots are treated as business-as-usual.

Academia will benefit through this project through the development of new planning models and algorithms capable of reasoning about mission risk given a dynamic, partially observable model of network communication, and the grounding of this approach in real industrial applications. Aims and objectives

- Establish the network conditions IMR robots will be operating under at nuclear sites in the future. - Determine an appropriate model for representing network availability to mobile robots. - Model the dependence on network availability for IMR actions. - Integrate the model into an algorithm for planning robot missions under uncertainty.

- Determine an appropriate risk metric to use as an objective function for this planning algorithm. - Extend the algorithm to plan for the movement of network access points. Novelty of the research methodology

This research develops novel planning models and algorithms capable of reasoning about mission risk given a dynamic, partially observable model of network communication. This approach fills a critical gap in current research, offering a new approach in managing mission risk in the presence of communication uncertainties.

EPSRC Research Area This project falls within the EPSRC Artificial intelligence and robotics theme. Companies Involved The Nuclear Restoration Services(NRS)/Magnox sponsor this project.