Adaptive Path Formation in Self-Assembling Robot Swarms by Tree-like Vascular Morphogenesis

For self-assembly, robot swarms can be programmed to form predefined shapes. 
However, if the swarm is required to adapt the assembled shapes to dynamic features of the environment at runtime, then the shapes' structures need to be dynamic, too. 
Prerequisite for adaptation is exploration and detection of changes followed by appropriate rearrangements of the assembled structure. 
We study a self-assembling robot swarm forming trees to explore its environment and searching for bright areas. 
The tree-formation process is inspired by the vascular morphogenesis of natural plants. 
Detecting light produces a virtual resource shared within the tree, helping to drop useless branches while reinforcing efficient paths between bright areas and the tree root.
We successfully verify our self-assembly approach in several swarm robot experiments in a dynamic environment showing that the robot swarm can collectively discriminate between light sources at different distances and of different qualities.