mocap_nav_demo

Navigation demo assuming perfect ground truth pose from motion capture system. Uses mocap localization feedback to drive Jackal robot along figure 8 path. Segmented cross-track error (CTE) error is used for PD control.

Dependencies

Required:

• ROS (need full desktop installation to test in Gazebo)
• vrpn_client_ros

Optional:

• urg_node (for stopping if an obstacle is in the way)
• pyqtgraph (for debug plot)

Robot

• Clearpath Jackal

Execution

On real jackal (assuming you are using a Jackal with laserscanner and it is running default bringup scripts and you have a motion capture system running a VRPN)

1. Add a tracker named "robot" to your motion capture VRPN

2. Launch motion capture client

roslaunch mocap_nav_demo vrpn_client.launch server:=<ip address of machine running motion capture server>

3. Launch laserscanner node (I am using Hokuyo urg-04lx. NOTE: this step can be skipped and the demo will still run, but make sure the robot has room to navigate and won't collide with anything)

roslaunch mocap_nav_demo hokuyo.launch

4. Launch mocap navigation demo for Jackal robot

roslaunch mocap_nav_demo jackal.launch 

On simulated jackal

1. Launch Jackal gazebo simulation with front laser (move robot to area where it has space to navigate)

roslaunch jackal_gazebo jackal_world.launch config:=front_laser

2. Launch mocap navigation demo for simulated Jackal

roslaunch mocap_nav_demo jackal_gazebo.launch

Control Pseudocode

1. S = segmented figure 8 shape
2. i=0 is current segment index
3. While True:
  a. Get robot pose from motion capture
  a. CTE = distance to current segment S_i
  b. ang_vel = PD_control(CTE)
  c. Publish twist command 
  d. If current segment is done -> i+=1