Footstep planner

This project is about:

• finding planes in the point cloud (with constraints on normals etc.)
• put footprints on proper planes to make a footstep plan for robot locomotion

Compile as a standard catkin_make project in your catkin workspace.

To use different robots you should put the packages containing their models in the catkin workspace, we are currently supporting two robots:

• walkman: package bigman_urdf in repo https://gitlab.robotology.eu/walkman-drc/iit-bigman-ros-pkg
• coman: package coman_urdf in repo https://github.com/EnricoMingo/iit-coman-ros-pkg

Startup of required software

roscore

Either specify a point cloud scene:

roslaunch footstep_planner fake_primesense.launch filename:=YOURTESTSCENE.pcd

or just load the default one:

roslaunch footstep_planner fake_primesense.launch

Start a static tf publisher in order to put a world frame into the point cloud, change the launch accordingly to the loaded scene.pcd. As an example, if you used stairs_2.pcd and you want to use the Walk-Man robot to plan you should launch:

roslaunch footstep_planner stairs_walkman.launch

Finally start the footstep planner:

roslaunch footstep_planner footstep_planner.launch

you can specify another robot with respect to Walk-Man using the robot parameter.

In order to send commands, just do a

rostopic pub footstep_planner/command_i std_msgs/Header TAB

And fill the automatic empty message that ros will put in the shell with any command below.

List of available commands

1. (cap_plan)

---

Reads a point cloud from a topic, recognizes planes, plans one step

2. (cap_save)

---

Reads a point cloud from a topic, recognizes planes, saves planes into a file

3. (load_plan)

---

Reads set of planes from a file, plans one step

4. (plan_all)

---

Plans as much step as possible until no steps are found

5. (draw_path)

---

Draws the planned sequence of steps up to now

6. TODO (direction x y z)

---

Sets the preferred walking direction (world frame?)

7. (exit)

---

Exit the footstep planner in a CLEAN way

Parameters of the point cloud vision algorithm:

• the grid size in meters for downsampling
• the search radius in meters for the normal estimation
• the curvature threshold for extract the planar areas in the downsampled cloud
• the minimum cluster size for euclidean clustering
• the cluster tolerance to accept points within a cluster in euclidean measure