(Experimentation for Objective 3)

In a software demonstration of a distributed robotics application where the robots must work together to complete an object placement task, will one robot be able to be trained from another robot (group-feedback) with similar performance to a robot trained by an ideal algorithm (“expert system”) for neurally-controlled navigation?

If complete environment information is available to the neural networks, assuming they consist of at least 4-5 layers and a sequential presentation algorithm is used, then it should be possible for an untrained navigation network to learn from a trained network. This would be an important step to present group-feedback as a viable possibility in distributed robotics.