Self-driving Modular AI-based Robot for Rough Terrain (SMARRT)

The Defense Advanced Research Projects Agency (DARPA) is seeking to utilize small reconfigurable robotic system to operate independently in unknown highly complex environments. We propose a modular reconfigurable robot, which presents several locomotion modality to handle different type of off-road environments. We proposed several configurations for the robot, which are well-suited to the type of terrain ahead of the robot. The proposed reconfigurable robot will be able to carry a load of 2-10 pounds over the course of 1 KM. We also propose a terrain classification and traversibility estimation algorithm to recognize terrain type and terrain degree of difficulty. The results of our proposed SMARRT system will be presented in simulation, where a simulated model of the robot as well as the physical environment will be used to show case the functionality of the system. At the end of Phase I effort, we will learn the advantages and disadvantages of such a reconfigurable robot when carrying on military missions.