Exploiting Behaviour Trees in Underwater Autonomous Robotic Networks

Abstract

Autonomy and cooperation are envisaged as the pillars for the effective control of robotic networks in the challeging underwater domain, strongly defined by limitations in communications. The development and the deployment at sea of the required advanced cooperative autonomy capabilities dictates that the adopted control architectures are characterised by modularity, scalability and ease of testing. Aiming at these objectives, we propose and describe the development of a mission manager module based on Behaviour Trees, integrated in our intelligent Cooperative Autonomous Decision Making Engine (iCADME) autonomy architecture. BTs are an AI technique, which provides reactivity, modularity and the appealing possibility to assemble complex missions by using sub-missions. All these features are central for the iCADME development, and increase the flexibility and adaptability from a mission execution standpoint. We report and discuss results of iCADME, working with a BT-based mission manager, during MEDASWAN23 trial. Our multi-robot underwater network was successfully controlled in a surveillance mission, involving real-time data fusion, cooperative behaviours, and the switching between different tasks. BTs resulted effective in simplifying the design of complex multi-task missions, easing their modification and their testing before the robot deployment.