ARCPro Ros2

arcpro waypointer example run

2 min read

Run commands

Run commands per sensor (and full examples) are located in 2 folders from content root:

  • drive_scripts:
  • hardware_launch_scripts:

Run with map

Each in a seperate terminal run the following:

  • colcon build and sorce
  • ./lidar.sh (in the hardware folder)

Lidar currently cant be brought up due to race condition (i think). Check nav2.yaml for the current map being used, they should be stored in the /res folder.

Run SLAM

We bringup in isolation, to build the lidar map we have the following all in the hardware folder:

  • vesc
  • lidar
  • joy
ros2 bag record -o name --topics /scan /tf /tf_static /odom

To replay and make the slam:

ros2 launch slam_toolbox online_async_launch.py slam_params_file:=/home/airou/PycharmProjects/arcpro_rl_base_case_waypointer/src/merger/config/slam_toolbox_params.yaml
 
# in another terminal
 
ros2 bag play name --clock

when finished run

 
ros2 run nav2_map_server map_saver_cli -f ~/my_new_map

See Slam Toolbox about some issues I ran into when attempting to run (due to odom configs)

Slam settings adjusted:

For the arcpro robots, I had to adjust the loop closure

  • file name: slam_toolbox_params.yaml

We use ceres_scan_matcher to match scans to map with slamtoolbox to create our initial map. We adjusted the settings to distrust the odometry as there was much issue with turning and keeping the same estimated pose (the graph would fracture like in Slam Toolbox).. We also vastly increased the occupied space weight so localizing within the /scan topic weight was increased. Loop closure was decreased slightly, but the minimum travel distance and orientation was vastly decresed to match the faulty odom topic. Given we were brute forcing the laser scan localization, we set these distances to be very short, and throttle scans to be quite low (at 2)

The package robot_localization was used to create our fused odom topic, which used the realsense IMU and motor encoders on our robot. We tried using pure IMU though it resulted in horrible x-y axis localization.

 do_loop_closing: true  
    loop_search_maximum_distance: 10.0  # (meters) How far to search for a loop closure.  
    minimum_travel_distance: 0.008        # (meters) Min distance robot must travel to add a new node.  
    minimum_travel_heading: 0.09  
  
 #me used this last time  
#    minimum_travel_distance: 0.05       # Was 0.1. Update every 5cm.  
#    minimum_travel_heading: 0.034       # Was 0.087 (5 deg). Update every 2 degrees.
    throttle_scans: 2                   # Process every 1 scan. Increase if CPU is too high.  
  
    solver_plugin: "solver_plugins::CeresSolver"  
  
    ceres_solver_options:  
      use_nonmonotonic_steps: true  
      max_num_iterations: 50  
      num_threads: 1

Graph View

Backlinks