Research - A Concept of Self-Repair for Robotics

If a human is injured, the biology of its body immediately begins repairing itself. Blood starts to clot in flesh wounds, antibodies are produced as a response to a virus, and neural structure changes to accommodate brain damage through neuroplasticity. Artificial neuroplasticity is the particular focus of this paper. In order to grant robotics the same liberties that humans enjoy in regards to self-repair, it will be shown how this can effectively be accomplished through the cooperation and reorganization of two neural networks. If a sensor sending input to an artificial neural network becomes damaged, the neural network of the damaged sensor will repurpose itself to be used for another task through self- reorganization and on-the-fly transfer learning. In other words, a robot can exhibit self-repair by borrowing from principles of biological neuroplasticity.