No description, website, or topics provided.
Clone or download
Fetching latest commit…
Cannot retrieve the latest commit at this time.
Permalink
Type Name Latest commit message Commit time
Failed to load latest commit information.
launch
src
CMakeLists.txt
README.md
package.xml

README.md

beta_adaptive_pHRI: Adaptive Physical HRI

Control, planning, and learning system for physical human-robot interaction (pHRI) with a JACO2 7DOF robotic arm. Learning is adaptive based on how relevant the human's interaction is.

Dependencies

  • Ubuntu 14.04, ROS Indigo, OpenRAVE, Python 2.7
  • or_trajopt, or_urdf, or_rviz, prpy, pr_ordata
  • kinova-ros
  • fcl

Running the Adaptive Physical HRI System

Setting up the JACO2 Robot

Turn the robot on and put it in home position by pressing and holding the center (yellow) button on the joystick.

In a new terminal, turn on the Kinova API by typing:

roslaunch kinova_bringup kinova_robot.launch kinova_robotType:=j2s7s300 use_urdf:=true

Starting the controller, planner, and learning system

In another terminal window, run:

roslaunch beta_adaptive_pHRI trajoptPID.launch ID:=0 task:=None method_type:=A record:=F replay:=F feat_method:=BETA feat_list:="table,coffee"

Command-line options include:

  • ID: Participant/user identification number (for experiments and data saving)
  • task: Task string {Bring close to table = table, Correct coffee cup orientation = coffee, Free roam = None}
  • method_type: Sets the pHRI control method {impedance control = A, impedance + learning from pHRI = B, demonstration = C}
  • replay: Replays a recorded bag of joint torques and angles
  • record: Records the interaction forces, measured trajectories, and cost function weights for a task {record data = T, don't record = F}
  • feat_method: Learning method used.
  • feat_list: List of features the robot's internal representation contains. Options: "table" (distance to table), "coffee" (coffee cup orientation), "human" (distance to human), and any combination of these.

References