Energy-efficient Tracking of Mobile Audio Sources via Emergent Distributed Systems

Tracking multiple mobile audio sources in acoustic scenes where the layout, targets, and requirements change rapidly is a fundamental problem in the research field of tracking only through listening with computational means. Recent developments in the field of robotics and artificial intelligence have enabled researchers to further the capabilities of such systems – interconnected or not – towards solving the localisation and tracking problem. Nonetheless, such research focuses primarily on managing the accuracy of such systems with little care for energy (i.e. battery) efficiency, especially in applications where movement is required. Meanwhile, highly dynamic acoustic scenes are not always accounted for in the designs using mobile listeners.

This thesis attempts to bridge these gaps by attempting to solve this problem with a focus on energy efficiency: reaching the targets in a timely manner conserving as much energy as possible. To achieve this goal a suitable system has been designed and implemented, while bio-inspired computing has provided the key inspiration towards developing a listening and tracking strategy that can expertly adapt to such scenarios. Established machine-learning techniques have been employed to further optimise this strategy, ultimately achieving even higher efficiency through adaptation of psychological research towards improved collaborative problem solving via emergence engineering.

The key contributions of this thesis are thus: a distributed system framework based on microservices tailored for modern devices capable of listening and tracking with both simulation and real-world deployment capabilities, an adaptive strategy that can be utilised for standalone system solutions, and an even more efficient approach for cooperative solutions. An example application could be the deployment of several small robots in disaster scenarios for reaching and aiding trapped individuals (e.g. building on fire with heavy smoke). Finally, the interdisciplinary research process followed throughout this undertaking aspires to offer an incentive for other researchers to pursue similar avenues for innovative applications or efficient solutions in the pertinent domains.