Scalar field guidance in swarms of simple robots

Developing swarms of robots as simple agents without any central controller is a challenging task. We focus on simple robots with limited capability which do not have a localization system to show them where they are. Termite construction activity inspires our work; although they have poor vision, they can construct a massive termite mound, build a royal chamber and communicate with each other by observing the changes in the environment, sensing vibrations and detecting chemical substances (pheromones). In our work, we used scalar fields as a global level of guidance for the robots. A scalar field associates a value with every point in the working region. We show how scalar fields can be used to guide a low-cost and limited-capability swarm of robots to execute a specific task. We present four examples of tasks using the scalar field, such as constructing shapes from ambient objects, finding the largest coverage-connected network among the robots, aggregating to a predefined area and foraging by finding and collecting ambient objects to the collection area. This work is divided into three parts; first, we show how the scalar field with different resolutions can help divide labour among the robots and guide them in their movements. Second, we investigate combining a scalar field with reinforcement learning to find the largest coverage network. Finally, we design a coloured scalar field and use it as a road network for the swarm to reduce spatial interference among the robots. We practically build our robots based on the Zumo robot kit to perform the aggregation task.