Path Generation Using Matrix Representations of Previous Robot State Data

Humans learn by repetition and using past experiences. It is possible for robots to act in a similar fashion. By representing past path traversal experiences with matrices, a new path can be generated without relying on calculations of complex dynamics or control laws. This paper presents one approach for allowing robots to use past experience to generate new paths and control actions. This approach relies on using several matrices to associate each new input value with previous robot states. An example is provided and analyzed which shows a successful simulated implementation of this approach. In addition a real world test of the approach was conducted which demonstrates that the implementation not only generates new paths, but does so fast enough to be feasible for real time systems