Project Manicopter: Multicopter-Based Robotic Arm for Aerial Manipulation

Transcription

1 Spacecraft Project Manicopter: Multicopter-Based Robotic Arm for Aerial Manipulation March 7, 06 Dr. Hyeongjun Park, NRC Postdoc Capt. Bruno Tavora, Ph.D. Candidate (Brazilian AF) PIs: Prof. Marcello Romano (NPS-MAE) Prof. Xiaoping Yun (NPS-ECE)

2 Contents. Introduction - Team - Motivation & Objectives. Multicopter model & parameter identification 3. Development of experimental platform - Robotic arm control through Pixhawk autopilot - Experimental setup 4. Summary

3 Team Staff Dr. Marcello Romano, Professor (MAE) Dr. Xiaoping Yun, Professor (ECE) Dr. Hyeongjun Park, NRC Postdoc Student Capt. Bruno Tavora, Ph.D. Candidate Collaborators Dr. Yoonghyun Shin, ECEP visiting scholar at NPS, ADD, South Korea Dr. Elisa Capello and Dr. Giorgio Guglieri, Politecnico di Torino, Italy Acknowledgement CRUSER funding Dr. Kevin Jones Mr. Steven Kuznicki, Pilot Engineering Group, Mathworks 3

4 Motivation & Background Motivation Physical interaction with the environment enables to extend utilization of UAVs to new type of missions Grasping, object picking & assembly, data acquisition and inspection by contact objects/surface Robust and stable maneuvers for aerial manipulation are challenging to achieve Background: Applications of aerial physical interacting Type Approach Application Aerial interaction Multiple UAVs connected to load Load transportation Air refueling & Device connecting two UAVs Refueling/Recharging Air-ground interaction Perching & Docking UAV with sampling device UAV with arm and gripper Refueling/Recharging Sampling Picking/Assembly Kondak et al., Unmanned aerial system physically interacting with the environment, 05 4

5 Spacecraft State of the Art Existing work by leading groups on aerial manipulation University of Pennsylvania (UAS) led by Prof. Kumar University of Seville (Spain) led by Prof. Ollero Seoul National University (South Korea) led by Prof. Kim - ETH Zurich (Switzerland ) led by Prof. Siegwart University of Bologna (Italy) led by Prof. Marconi University of Twente (Netherlands) led by Prof. Fumagalli Pick & Place Multi-UAV load transportation Assembly tasks Mainly contact with objects Contact with vertical surface Surface inspection 5

6 Research Objectives Investigation of the dynamics, guidance, and control of autonomous multicopters with robotic manipulation capability Development of an experimental platform of a multicopter with a robotic arm Analysis and experimentation for the system dynamics when the multicopter contacts with the environment Development of attitude controller and guidance algorithm for real-time path-planning to contact and avoid obstacles Implementation of mission scenarios Picking & Assembly Door/Drawer opening Data acquisition on surface 6

7 Scenario Simulation Video. 7

8 Model Description: Multicopter Hexacopter 6,C,A Six motors and propellers Six arms connected symmetrically 5,A,C X Y " " 6 3 4,C 3,A 5 4 Z B Part Description Weight [g] Autopilot 3DR Pixhawk 39 Electric motor T-Motor KV X 6 Propeller E-Prop 54 X 0 X 6 LiPo Battery & Power part Thunder power 800 mah 69 C: Clockwise A: Anticlockwise Total weight 89 Payload weight 000 8

9 Modeling: Nonlinear Model Nonlinear model of the hexacopter! φ p qsinφ tanϑ r cosφ tanϑ, ϑ! qcosφ r sinφ, qsinφ r cosφ ψ!, cosϑ cosϑ u! qw rv g sinϑ, v! w! p! q! r! pw ru g cosϑ sinφ, pv qu g cosϑ cosφ I xx I yy I zz m ( T T T T T T ) [( T T T T ) d ( T T ) l qr( I I ) I qω], 4 [( T T )( d d ) ( T T )( d d ) ( T T ) d pr( I I ) I pω], [( Q Q Q Q Q Q ) pq( I I )] p 3 3 xx 4 yy yy Unknown parameters 5 zz 5 6 3, p Angles: φφ, θθ, ψψ Angular velocity : pp, qq, rr Translational velocity : uu, vv, ww zz xx YY XX % % 6 p 5 Moments of inertia Thrust of motor Angular velocity of propellers 4 3 ll ( Motor reaction torque 9

10 Parameter Identification It is important to understand and obtain a precise model of a multicopter for advanced controllers Direct computation of geometry [Chovancova et al. 04; Elsamanty et al. 03] Analysis from flight data [Stanculeanu et al. 0; Chovancova et al. 04] We devised and implemented an identification method: Compound pendulum method [Gracey, NACA Technical Report, 948] Evaluation of principal moments of inertia and engine thrust Vicon motion capture system Infrared marker-tracking system with millimeter resolution for position and attitude of hexacopter 0

11 Experiment on Moments of Inertia Video.

12 Identification of Moments of Inertia Lagrangian approach ( ) ( ) θ θ θ )cos ( ) cos (, ) (, d l mg l m g V I I d l m m l K V K L rod! ( ) ( ) ( ) θ θ θ θ θ θ θ θ θ B d l mg m gl d l mg m gl V A I I d l m m l K dt d V K dt d rod ) ( sin ) (, ) ( 0,!!!!!! I I d l m m l d l mg m gl A B B A rod n ) ( ) ( 0, ω θ θ!! θ l d l Yaw

13 Identification of Moments of Inertia Period of oscillation T I p π ω T n p 4π π B A, ( mgl mg( l d )) ml m( l d ) Irod All known values except period TT ( Period TT ( is measured by Vicon system (analysis of time history) Yaw angle Pitch angle Principal moments of inertia II [kkggmm? ] II AA [kkggmm? ] II EE [kkggmm? ] 3

14 Identification of Engine Thrust Thrust generated by a motor and propeller Relation between PWM signal and rotational speed of a rotor is known Thrust PWM RPM Thrust experiments: Thrust vs. PWM input to a motor Use the compound pendulum and Vicon system One propeller 4

15 Experiment on Thrust Video 3. 5

16 Identification of Engine Thrust Thrust experiments m gl sinθ mgl sinθ Tl 0 l θ Different experiments are performed with different PWM inputs PP to a motor, 00 PP 900 [μμss]. T 0.005P l l mg mg T Thrust vs. PWM RPM vs. PWM 6

17 Identification of Torque Relation of aerodynamic torque produced by motors and propellers Ad hoc method using a floating test bed & robots The robot with the hexacopter is floating on the granite rig Three propellers rotating in the same direction are mounted Dimension Surface precision Leveling precision Planar accuracy 4mm 4mm 0.3mm AAA < 0.0 ddeegg ±0.007 mmmm - Linux Real-Time workstation - Ad-Hoc WiFi internal network 7

18 Experiments on Torque Video 4. 8

19 Identification of Torque Experiments on the floating test bed Two cases and comparison (torque is assumed to be constant) Case : No propellers on & 4 thrusters of the floating robot on for 5 s Case : Three propellers on & No thrusters of the floating robot on Torque ττ II U ωω E In Case, ττ is known and equal to [NNmm] and ωω E is measured from Vicon. Hence, II E is obtained Relation between PWM and torque: Measured yaw angle τ 0.000P Torque vs. PWM

20 Validation by Flight Data Flight experiments to validate the results of parameter identification Initially, the hexacopter hovers on a desired position, and performs accelerating movement Verify the nonlinear model and parameters Experimental setup 0

21 Flight Experiment for Validation Video 5.

22 Experimental Results Generation of angular acceleration Hexacopter hovers Impulsive inputs are applied to generate pitch acceleration at 0. s and. sec Estimation is overall well-correlated to measured values Pitch acceleration Yaw acceleration

23 Improvement of Test Platform Mission planner APM firmware Embeded PID attitude controller Limitations 3

24 Previous Experimental Setup Mission planner APM firmware Ground Station Vicon System Hexacopter 4

25 Previous Experimental Setup: Issues Delay of 0.4 sec Resolution: Each command can be quantized among 53 different values only APM firmware can only run a specific PID controller, whose gains are the only parameters that can be changed No capability of sending and receiving raw data through serial ports. However, it s essential to get access to the serial ports to have communication between Pixhawk and Arbotix- M, the robotic arm controller board. 5

26 Simulink Pixhawk Support Package Opensource Customized communication blocks provided by Pilot Engineering Group, Mathworks 6

27 New Experimental Setup with Simulink Ground Station Hexacopter Vicon System Robotic Arm 7

28 New Experimental Setup with Simulink Small delay and better resolution More flexibility to attitude controller design, not more restricted to a PID controller Pixhawk and Arbotix-M can be connected through a serial cable 8

29 Control Robotic Arm through Pixhawk Video 6. 9

30 Summary Development of an experimental platform for a multicopter with a robotic arm (achievement since May, 05) Modeling and parameter identification via compound pendulum and Vicon system Improvement for communication and attitude control algorithm development Robotic arm controlled by Simulink Package for Pixhawk Future work Flight experiments with the robotic arm Development of attitude controllers Study on collision detection/response using the robotic arm Real-time path-planning to avoid obstacles 30

31 Spacecraft Thank you