In this video you can see how the UAVs and rover acts in a simulation environment.
Background
The main goal of the project is to develop a search and track system consisting of an autonomous tracked vehicle called
a rover and two low-altitude autonomous UAVs, Unmanned Aerial Vehicles. The rover is responsible for mapping the
environment, using a LIDAR, and searching for intruders using a camera. The UAVs only search for intruders using
cameras. Once the intruders have been found, they will be followed by the rover and one UAV until the capture is
considered complete. Once completed, the rover and UAV will continue searching until all intruders have been found
and the mission is complete.
System description
The Search and Track project consists of two UAVs and a rover that communicates through a base station.
The UAVS utilises an RRT* motion planner with straight line path segments in 3D space. This algorithm is asymptotically optimal and ensure that the UAVs can actually follow the planned path. The motion planner also plans around so called no-fly zones. An MPC controller with a 3D space model for the position of the UAVs was used for path following with velocities in each direction as control signals. The UAVs are equipped with a multitude of sensors, mainly within the flight controller, such as accelerometers, gyroscopes, barometer and magnetometer. However, external positioning was used for better accuracy. They are also equipped with a camera that was used to identify intruders which utilized an EKF with a camshaft algorithm.
Rover
The rover utilises an RRT* motion planner with Dubins path segments. This algorithm is asymptotically optimal and ensure that the rover can actually follow the planned path. An MPC controller with a differential track speed model of the rover was used for path following. The model was transfered to the Frenet frame to decouple the two error states of the rover as much as possible. The rover is equipped with a multitude of sensors, including a camera, an IMU, wheel speed sensors and a LIDAR. The camera was used to identify intruders and used an EKF with a camshift algorithm. The other sensors were used for positioning and mapping and also utilised an EKF.
Base Station
The base station module is responsible for the communication between the agents and the structure of the mission. A search algorithm is used which implements a probability map of the area that updates the possible location of intruders using information received from the agents. The mission planner is modularly built and designed to be independent of the amount of agents used for the mission.