The main objective under consideration is to build a platform consisting of heterogeneous systems for the efficient localization and path following of multiple ground robots in the world frame of reference. The idea is to use a downward facing camera attached to a quadrotor UAV at hover, to determine the position of the ground robots with respect to the quadrotor. The position of the quadrotor with respect to the world frame is assumed to be known. Any inaccuracies in quadrotor's position will be reflected in the position of ground robots with respect to the world frame. However, the positions of the ground robots with respect to each other will remain unaffected.
PD controllers are designed for ground robots to follow respective waypoints in the camera view. The leader robot is directed to follow a circular path with a series of waypoints. The PD design objective for follower robots is to maintain a certain distance and angle from the corresponding robot at front. This distance and angle can be converted to the desired cartesian coordinates, in leader's frame of reference. This means that the follower holds a fixed position with respect to the leader. Moreover, every follower is considered to be holding this fixed position with respect to the robot in front of it.
Simulations for the project are performed in ROS/Gazebo. Transform publishers are used to publish respective transforms. The operator node on the quadrotor listens to the world to leader, leader to follower and follower to follower transforms, and publishes them. Leader robot subscribes to the topic associated with world to robot transform, and follow the desired waypoints. Follower robot subscribes the leader to follower transform and try to maintain that position. Robot behind a follower robot does the same considering that the next follower is its leader. In the experiment performed, we consider the follower to maintain 0.3 meters distance and 20 degrees angle from the leader. Whereas, the leader is expected to follow a circular path of radius 0.4 meters. Three ground robots are included in the simulation. Figures below are directly exported from ROS/Gazebo. Figure 1 shows x and y coordinates of position for the leader robot with respect to the world frame. Figure 2 shows the position of the first follower in leader's frame of reference, whereas Figure 3 shows the position of the second follower with respect to the first follower. All the graphs are plotted against time.
This is a simulated demonstration of the testbed that we have achieved so far. The control parameters must be tuned; however, the video demonstrates the capabilities of the testbed. An aerial quadrotor with a downward facing camera is able to provide localization information for the ground agents. This information is used by the ground agents for path planning and control feedback purposes.