Many years ago, I worked with the University of New Mexico and NASA on mobile robot swarms to efficiently collect samples from Mars on a foraging expedition. Given cheap robots built at $2000 a pop with inexpensive parts such as a camera, sonar sensor, gyroscope, an open-close arm, and Arduino setup with ROS, we endeavored to build and train the robots to forage for resources.
These robots do not have a map of the terrain. Like ants, in their search for food, they wander randomly in the classic drunken salesman problem. The swarmies only understand their local maps with respect to a central place (the nest) to guide them. There are only four states in swarmie world: search for target, retrieve target, deliver target to home, and avoid obstacle. Our swarmies only had 2D representation of the world, but we used the camera to identify targets and triangulate for depth perception. They could determine how far away the identified resources were to move toward them.
Ants release pheremones in order for others to follow back to the ant hill nest. Since the swarmie robots don’t have pheremones, we had swarmies return to the nest if they were in search mode to follow the successful swarmie to the cluster of resources it found by following the movement of the April tag on the back of each rover. Once each swarmie knew where the cluster of resources were in their local map, they could all work together to collect the known resources and return to searching for new sources.
This could be applied to many systems scenarios with these cheap but effective swarms to increase yield. For example in agriculture, a swarm of robots goes in to till the fields and another swarm plows the soil. When the time is right, the swarm goes in to plant seeds. These swarms perform tasks that require basic sensors to perform fundamental and simple tasks. Similarly, swarms could be effective in search and rescue missions.
Perhaps eventually, when robot swarms are sent to Mars, they could even build a home base with resources from the planet itself. Or duplicate themselves using materials found.
Stranger things are apt to happen, right?