❓ State Estimation
The state of the SMB can be estimated in multiple ways: using the tracking camera, fusing LiDAR + IMU using Graph MSF. Then, launch the simulation environment in planner_tutorial.world:
Using Tracking Camera
The SMB robot is equipped with an Intel Realsense T265 tracking camera, which can output the odometry using visual-inertial odometry. To open the simulation with the tracking camera, run the following command:
# In the host PC
roslaunch smb_gazebo sim.launch tracking_camera:=true launch_gazebo_gui:=true keyboard_teleop:=true
This will open up Gazebo and RViz. In RViz, change the fixed frame to tracking_camera_odom, then drive the robot with keyboard controls and see how the robot is moving in RViz respective to the odom frame.
Using LiDAR + IMU
The SMB robot is equipped with LiDAR and IMU. For state estimation, these data can be fused using the method Graph-MSF: Graph-based Multi-sensor Fusion for Consistent Localization and State Estimation. To learn more, link.
To build MSF Graph,
catkin build -c smb_msf_graph
To run the MSF Graph,
# in one terminal - launch Gazebo and RViz
roslaunch smb_gazebo sim.launch launch_gazebo_gui:=true keyboard_teleop:=true
# in second terminal - launch msf graph
roslaunch smb_msf_graph smb_msf_graph.launch use_sim_time:=true
If successful, you can see GMsf ...graph is initialized. message in the second terminal.
In RViz, select world_graph_msf as the global frame. Then drive the robot with keyboard controls and see how the robot is moving in RViz respective to the world frame.