Optimized distributed scheduling for a fleet of heterogeneous unmanned maritime systems

Due to the increase in embedded computing power, modern robotic systems are capable of running a wide range of perception and control algorithms simultaneously. This raises the question where to optimally allocate each robotic cognition process. In this paper, we present a concept for a novel load distribution approach. The proposed methodology adopts a decentralised approach towards the allocation of perception and control processes to different agents (unmanned vessels, fog or cloud services) based on an estimation of the communication parameters (bandwidth, latency, cost), the agent capabilities in terms of processing hardware (not only focusing on the CPU, but also taking into consideration the GPU, disk memory speed and size) and the requirements in terms of timely delivery of quality output data. The presented approach is extensively validated in a simulation environment and shows promising properties.