Experimental Investigation of a Semi-Autonomous Search and Rescue System

A previously-proposed semi-autonomous search and rescue system is investigated experimentally. A stationary vision system identifies and tracks a victim using an object detection neural network. The vision system then guides an autonomous unmanned ground vehicle to the victim's location using a ROS 2 publisher-subscriber communication protocol. A series of experiments are conducted with off-the-shelf components to investigate the impact of several design parameters. The resulting experimental data demonstrates that a fast and accurate response can be achieved. Furthermore, guidelines are established for parameter selection in future experiments.