Creative Mobile Terrain Sensing Multi-Valued Behavior Robots

Within our Phase I effort, a radically new kind of neural network based control system that broadly emulates the contemplative capacity of the human brain has been successfully tested aboard a complex hexapod robot. In the course of this preliminary investigation, this neurocontrol system was able to bootstrap its own locomotive behaviors on demand, integrate camera images and sonar returns to build navigational potentials, and calculate paths of least resistance to pre-designated targets through its surrounding terrain. In the proposed Phase II effort, we will package these highly adaptive control paradigms through the use of a class factory that can readily tailor object-oriented neurocontrol systems for a wide range of wheeled, tracked, and legged robots. Wherever possible, we plan to use hardware acceleration to increase computational speed, striving to place this computational intelligence directly onboard the robot. To prove the flexibility of the generated control classes, we will instantiate them on a range of mobile robots: (1) NASA’s omni-wheeled Robotics Technology-Demonstration Platform (2) AFRL’s treaded ARTS platform, and (3) a family of complex hexapod crawlers. In the final stages of this program, we will tailor each of these robots to specific commercial applications, using this highly flexible control class factory.