Task scheduling is a key challenge encountered in many distributed computing applications that involve multiple computing nodes. With the growing popularity of Artificial Intelligence, robots have been widely used in many commercial and military applications such as environmental monitoring, search and rescue, warehousing and logistics, etc. In this project, we aim to develop an efficient task scheduling algorithm for multi-robot systems with each robot being a computing node. Although task scheduling has been extensively studied, existing approaches mostly consider systems with static computing nodes and are not scalable to the number of tasks. Nevertheless, robots can move, which may cause frequent topology changes, packet losses, and time-varying communication delays. To address these unique characteristics of the multi-robot system into consideration we propose an online and adaptive Graph Convolutional Network based task scheduler, which generates task schedules dynamically according to the state of the network. It can be trained using any existing scheduling algorithm as a teacher and produce results that are similar but faster than the teacher
