• Gantry playback of UAV motions
• Insertion experiments

• Test 2 - Gantry mimicking UAV motions with Cartesian-controlled end-effector performing insertion task with compliance

• Test 1 - Gantry mimicking UAV motions with Cartesian-controlled end-effector performing insertion task

• Gantry playing back log file of x-y-z positions and velocities of UAV flight

• Position history (x-y)

• Conference paper writing (ICRA)
• Gain tuning

• Controller testing without manipulator attached

• Marker tracking and arm articulation

• ICRA 2014 Video Submission - Peg-in-hole testing, flight tests with arm articulation, z-control, and marker tracking

• 3-DOF Arm, 1-DOF gripper, manual control

• First flights - manual control, tethered power and communications, ROS infrastructure using mavlink and roscopter, 3 serially linked MX-28 Dynamixels implementing a spherical wrist (yaw, pitch, roll). Arm left un-actuated (not powered).

• Thesis writing
• MM-UAV aircraft and arm development

• Progress toward a fast-IK solver
• Simulation environment and testing

• Simulink Block Diagram

• Simulink Animation

• Pick-and-place and peg-in-hole at the Larics Lab

• Robotics Toolbox simulation model for MM-UAV

• Visual servoing using motion capture and an ARTag marker

Video 1 - End-effector tracking a motion capture marker

Video 2 - End-effector tracking an ARTag

• Joint work with Matko Orsag to tune MM-UAV velocity and pitch controller

• Peg-in-hole insertion using Cartesian impedance control

Video 1 - No UAV reaction

Video 2 - Contact force test with UAV reaction

Video 3 - Insertion and missed insertion with UAV reaction

• Control Architecture

• Applied pitch torque and UAV linear velocity

• Presented paper at ICRA conference in Karlsruhe, Germany
• Presented paper at ICUAS conference in Atlanta

• Picture - ICRA presentation

• Presented paper at TePRA conference in Boston
• Built ROS Moveit package for MK2 Arm

• Video1 - Moveit motion planning test

• Final TePRA paper submitted for publication
• Final ACC paper submitted for publication

• Video1 - Tool usage tests

• Video2 - Improved MM-UAV Gantry control through torque feedback
  • The torque sensors in the MK2 torso measure applied joint torque which is then added as a disturbance to a linear velocity reaction in the gantry

• Specifics on [http://www.hdtglobal.com/services/robotics/mk2-robotic-arm/ HDT MK2 Robotic Arm]
• TePRA paper accepted for publication
• ACC paper accepted for publication
• ICUAS paper submitted
• Conducted torque profiling experiments
• Attended conference call between ARL-DU-Drexel on MM-UAV collaboration

• Video1 - MM-UAV Gantry control through torque feedback
  • The torque sensors in the MK2 torso measure applied joint torque which is then mapped to a linear velocity reaction in the gantry
  • Impedance control is currently enabled to allow for a push-pull on the aircraft

• Images
  • Hose insertion testing and torque profiling

Approach
Movement
Contact

• Torque profile for elbow pitch joint

• Video2 - MM-UAV tool usage
  • Arm easily grabs and drills a hole. Joint control is difficult through a joystick which is why the movements are jerky and slow

• Video3 - MK2 impedance control testing
  • Elbow pitch and wrist pitch joints set with inertia, stiffness, and damping parameters

• Video4 - MK2 RViz Simulation

• ROS-packages operational
  • /command messages and /joint_state feedback for 3-DOF torso, 7-DOF arm, and 4-DOF end-effector
  • Gantry control through /cmd_velocity messages
  • Gantry feedback through /tf published by motion capture system
  • CAD exported to URDF with RViz mimicing joint states
• Mounted HDT MK2 Arm and Torso to 3-DOF Gantry
  • Arm consists of 7-DOFs to include shoulder pitch, roll, yaw; elbow pitch; wrist yaw, pitch, roll joints
  • 4-DOF end-effector has opposable thumb yaw joint, thumb pitch joint, index finger pitch joint, and ring finger pitch joint
  • Torque feedback on all joints. Finger actuators provide ~2 Nm of torque
• TePRA paper submitted
• ICUAS paper in-progress

• Video1 - MM-UAV grasping test
  • End-effector easily provides enough grasping force for 1600g hose. Arm is joystick-operated to insert hose into pump.

• Video2 - ROS and Rviz environment
  • ROS-package provides control of each actuator and sensor feedback updates robot and joint states in Rviz.

• Still Images
  • Concept poses for potential MM-UAV tasks (all poses were done through tele-op control)

• All actuators can be commanded with position/velocity/torque feedback. Impedance control has been tested but not implemented. Motion capture is operational. Gantry control through ROS.
• Software infrastructure consists of ROS and OpenRAVE

• Develop an ikfast plugin using OpenRAVE
• Run a whole-body kinematic trajectory on MM-UAV
• Implement aircraft model and attitude controller on gantry and gimbal
• Implement impedance control on manipulator to provide active compliance during task execution

MM-UAV will focus on three classic control problems: peg-in-hole, value turning, and door opening. These task align with the DARPA Robotics Challenge Events 4, 7, and 8.

The first task under study is Task 8.

• Task 8 - Connect cable or hose
  • Perception ability to locate and manipulation ability to make connection
  • Hose could be firefighting water hose or electrical cable
  • Will have to carry across terrain then connect

• Hose installation
  

To complete the task, things listed below must be integrated.

• Grasping
• Manipulation
• Lifting & Carrying
• Flight Stability
• Impedance/compliance control

• Drexel Autonomous System Laboratory (DASL) directed by Dr. Paul Y. Oh, Mechanical Engineering and Mechanics, Drexel University
  • Kinematic & Dynamic motion planning for whole-body lifting and carrying
  • Whole-body motion control design

Contact info: Paul Y. Oh, Professor: paul@coe.drexel.edu, Christopher Korpela, Ph.D candidate: cmk325@drexel.edu