An energy-aware architecture : a practical implementation for autonomous underwater vehicles

Abstract

Energy awareness, fault tolerance and performance estimation are important aspects for extending the autonomy levels of today’s autonomous vehicles. Those are related to the concepts of survivability and reliability, two important factors that often limit the trust of end users in conducting large-scale deployments of such vehicles. With the aim of preparing the way for persistent autonomous operations this work focuses its efforts on investigating those effects on underwater vehicles capable of long-term missions. A novel energy-aware architecture for autonomous underwater vehicles (AUVs) is presented. This, by monitoring at runtime the vehicle’s energy usage, is capable of detecting and mitigating failures in the propulsion subsystem, one of the most common sources of mission-time problems. Furthermore it estimates the vehicle’s performance when operating in unknown environments and in the presence of external disturbances. These capabilities are a great contribution for reducing the operational uncertainty that most underwater platforms face during their deployment. Using knowledge collected while conducting real missions the proposed architecture allows the optimisation of on-board resource usage. This improves the vehicle’s effectiveness when operating in unknown stochastic scenarios or when facing the problem of resource scarcity. The architecture has been implemented on a real vehicle, Nessie AUV, used for real sea experiments as part of multiple research projects. These gave the opportunity of evaluating the improvements of the proposed system when considering more complex autonomous tasks. Together with Nessie AUV, the commercial platform IVER3 AUV has been involved in the evaluating the feasibility of this approach. Results and operational experience, gathered both in real sea scenarios and in controlled environment experiments, are discussed in detail showing the benefits and the operational constraints of the introduced architecture, alongside suggestions for future research directions.