ksatyaki.github.io

MoD-aware motion planning benchmark

• MoD-aware motion planning benchmark
  • MoD-building repos
    • Instructions and examples
  • MoD-planning repos
    • mod_ros
    • ompl_planners
    • mod_path_planning
  • Multi-agent coordination based Simulation

MoD-building repos

1. CLiFF-map is available here.
2. STeF-map is available in the mod_ros repository below.
3. Intensity-map and GMMT-map is available in the useful_scripts package.

Instructions and examples

Some programs such as the gmmtmap.py in useful_scripts require that the input data file is in the ATC format (the same format used in the ATC dataset), but ordered by trajectory id. The useful_scripts package contains many programs that help us convert files to the ATC format, or cut out pedestrian data at specific times, etc. For example, the split_into_segments.py is used as follows:

• Data generated from PedSim is available as a ROS-bag file. Let’s call this data1.bag
• This data is converted to ATC format using spencer_rosbag_to_atc.py → data1.csv
• We need to order this data using the id of the object. So we run order_by_id.py → data1_ordered.csv
• PedSim tracks all its pedestrians continuously, so we can set a border (outside which we pretend that PedSim loses the pedestrian) and use this border to split the different trajectories appropriately.
• For this we run split_into_segments.py that produces the desired result.

Notice that the above example is only if we need to generate a GMMT-map from the data.

MoD-planning repos

mod_ros

• ROS packages that read and publish different MoD maps to ROS services.
• ROS Messages and Service support for GMMT-map and Intensity-map (available here under atc_dataset).
• ROS Messages and Service support for CLiFF-map (available here).
• ROS Messages and Service support for STeF-map available in this repository including ROS interfaces.

ompl_planners

• Libraries providing Optimization Objectives (ompl) based on all the MoDs in mod_ros.
• This is used in conjunction with the mod_path_planning

mod_path_planning

• Contains pre-built MoDs and their occupancy maps for certain environments.
• Planner executable and launch files.
• Please look at the launch directory and especially the files: planner_pedsim_warehouse.launch and planner_atc.launch.
• Please also look at the tmule directory, and especially this file.
• We recommend using TMuLE. It’s super easy to install via pip and makes life a lot easier.
• The TMuLE scripts basically tell tmux to launch certain scripts/programs in certain tmux windows and tabs. So these are all the programs one needs to start in order to start the motion planning.
• The last window in the tmule yaml file is the list of testing scripts (these post messages via ROS with the start state, goal state, planning time, etc.).

Multi-agent coordination based Simulation

• Repository available here.
• For our simulation we need the uncontrollable-robots-pedestrians branch!
• Take a look at the directory here
• It contains the MultiplePedestriansAndRobot.java file and other scripts that make it easier to run several tests together. :-)
• Don’t forget to contact me via github if you have any questions!

Data used in the experiments:

• Paper 1 under https://gitsvn-nt.oru.se/ksatyaki/mod_journal_stats