Modular Design and Implementation for Rapid Deployment of Autonomous Systems

This paper describes the autonomous systems, which Technical University of Denmark used to participate in Challenge 2 of the Mohamed Bin Zayed International Robotics Challenge. We participated with two autonomous vehicles in the Challenge: an aerial and a ground vehicle. The mission of both of the vehicles was to locate blocks and use them to build a wall in a marked location. Our ground solution consisted of a SKID steered vehicle, with a Universal Robots arm attached to it, and our aerial solution was a DJI M100 quadrotor in X configuration, equipped with a rangefinder and camera. Both platforms each have their own custom build end-effector, designed for lifting flat magnetic objects. The software was designed with a modular approach based on the mobotware framework, such that mission scripts could rapidly be assembled at the deployment site. A state of the art neural network, for detecting blocks, was trained for our ground vehicle. The effectiveness of the modular approach was tested in the challenge, and our lessons learned is included in the paper.