The most important issue in designing a self-reconfiguration planner is defining the metric that indicates the difference between two configurations. For most two-dimensional lattice systems and for isotropic three-dimensional modules, there is a good correspondence between the lattice distance.
Along with this problem, some generic hardware constraints must be taken into account.
These photos are taken from video. The first frame is a walking configuration.
A neural oscillator which drives the modules, and a GA, which optimizes the neural oscillator network.
Sort the individuals by their fitness and perform selection, crossover and mutation procedures.
Go to Embodied Artificial Intelligence - Part 1
Go to Embodied Artificial Intelligence - Part 3
Main page for Science and Engineering