Gestures-teleoperation of a heterogeneous multi-robot system

This work presents a solution for the teleoperation of a heterogeneous team of mobile robots. Regarding the team of robots, two possibilities are considered which are UAV-UGV (Unmanned Aerial Vehicle-Unmanned Ground Vehicle) and UAV-UAV. To execute this task, high-level gesture patterns are made in a remote station, and we proposed an easy-to-train Artificial Neural Network (ANN) classifier to identify the skeletal data extracted by an RGB-D (Red, Green, Blue-Depth) camera. Our classifier uses custom data to build the gesture patterns, allowing the use of smooth and intuitive gestures for the teleoperation of mobile robots. To validate our proposal, experiments were run using two off-the-shelf Parrot AR.Drone 2 quadrotors and the differential drive platform Pioneer 3-DX. The results of such experiments allow concluding that the proposed teleoperation system is able to accomplish inspection/surveillance tasks, and it can be easily modified to similar applications, as emergency response or load transportation.