Experimental Validation of Stable Obstacle Climbing with a Four-Wheel Mobile Robot OpenWHEEL

Transcription

1 al Validation of Stable Obstacle with a Four-Wheel Mobile Robot OpenWHEEL i3r Jean-Christophe.Fauroux@ifma.fr Belhassen-Chedli.Bouzgarrou@ifma.fr Frederic.Chapelle@ifma.fr Clermont-Ferrand, France LaMI UBP IFMA TIMS Mechanical Engineering Research Group Blaise Pascal University French Institute for Advanced Mechanics Research Federation 1

2 More agile mobile robots needed in the future Spatial exploration Mars exploration robot A rover for sample analysis Challenging usual applications Wheeled ATVs are blocked on ground discontinuities Wheelchair blocked before an obstacle 2

3 One big machine Agriculture A fleet of agile robots Rescue Fleet of robots to avoid soil compaction Earthquake Mag. 7.4 Turkey (1999) Scanning the streets and buildings Beach cleaning uses big machines or manual cleaning Beach pollution Towed filtering machine Manual cleaning 3

4 This work is about: Wheeled robots That climb step obstacles With only four wheels And stable behaviour Within the OpenWHEEL framework Wireless connection A3 Rear S 32 S 31 Wheel W31 A2 I2 S 21 W21 Double wishbone W22 Innovative suspension S 11 W11 S 12 A1 I1 W12 Camera S22 Suspension mechanism Saw Swing arm CAN Bus W 32 n Fro t Z X Y Inter-axle mechanism Ia Serial Parallel Innovative mechanism mechanism mechanism 4

5 Agile mobile robots Terrestrial locomotion system - unilateral / bilateral - slipping / sticking - can change in nature and number Terrestrial vehicles & robots Poly-articulated mechanical system Interact with environment Contacts with the ground Wheeled vehicles prevail (energetic efficiency?) Blocked on slope discontinuities of the ground Legs / Tracks regain interest for climbing Interface with the ground Crawler + multiples contacts, can cross obstacles & rough terrain - require high energy, moderate speed, complex control Leg + can cross obstacles and go fast on rough terrain - contact discontinuity, energy cost, stability control Wheel + fast on smooth surface, energy efficient - cannot climb obstacles or run on rough terrain Track + permanent stability, high traction - high friction energy loss, particularly during steering 5

6 Mobile robots based on legs Bi / Quadri / Hexa / Octo Natural gait / self-teaching Legs with feet = wheels Yanboo III (13kg,0.7m high) Biped with suction/rolling effectors Legs are manipulators www-robot.mes.titech.ac.jp Leg Wheel Hybrid Big Dog (75kg, 1m long, 6km/h, 35 slopes, 150kg payload) Roller-Walker (24kg, 0.5m long) Convertible wheels / Dual locomotion mode: walking / roller-skating www-robot.mes.titech.ac.jp 6

7 Wheeled robots Wheel: energy efficient even when steering Only exception : skid steering Leg Wheel Hybrid Pioneer P3-AT Skid steering simple robot Nomad Dual Ackermann steering strategy 7

8 Adaptative Wheeled Robots Minimally actuated frame, energy efficiency Simple control Leg Wheel Hybrid Micro5 abilities via 5 wheels Rocky 7 Adaptative rocker-bogie structure www-robotics.jpl.nasa.gov 8

9 Adaptative Wheeled Robots Crab I Adaptative parallel bogies Obstacle climbing abilities Leg Wheel Hybrid Shrimp 6 wheels on 2 // bogies and 1 front linkage 9

10 Hybrid multi-mode robots Highly actuated frame Orientable tracks for special modes of displacement Leg Wheel Hybrid Azimut 4 orientable tracks Helios VII 2 articulated tracks + 1 manipulating arm with hybrid grip/wheel end effector www-robot.mes.titech.ac.jp 10

11 Hybrid multi-mode robots Leg Wheel Hybrid Highly actuated frame Displacement modes: peristaltic crossing, obstacle climbing RobuROC 6 (150 kg, 1.5m long) 3 tiltable axles with passive warping Able to turn on itself Can climb obstacles Hylos (0.5m long) 4 wheels on 3DOF legs Lama Peristaltic crossing of sandy areas 11

12 Chosen locomotion modes for OpenWHEEL Chosen architecture OpenWHEEL OW i3r Stability Rolling step obstacles Multi-modes Only four wheels Hybrid robot Two locomotion modes (at least) Wheels (efficiency) On legs (climbing ability) No existing robot that climb with only four wheels A challenge for stability during climbing Easier to transfer on real vehicles Few actuators Actuated wheels Few actuators in legs/frame for better simplicity / stiffness / consumption / price 12

13 OpenWHEEL i3r i3r = 3R inter-axle mechanism 2 passive R joints in the middle of the axles 1 actuated R joint for central warping Kinematic structure with double symmetry OpenWHEEL i3r A big central actuator for warping Passive joint Axle steering without robot motion OpenWHEEL OW i3r Stability ar e R W21 W22 nt o r F W11 W12 Four actuated wheels More space for payload 13

14 Stability criterion on 3 wheels 2D modelling when lifting one wheel Wheel W12 (front-left) Wheel W21 (rear-right) 1) 2) 3) Front axle steering Wheel W11 (front-right) W12 4) W11 G G G OW i3r Stability 5) Rear axle steering OpenWHEEL W22 G W21 Wheel W22 (rear-left) W11 G 6) W12 7) G 8) G G W21 W22 Stable Unstable Stable Unstable Stability if the lifted wheel is inside the turn 14

15 Stability during climbing 3D modelling Stability margin on 3 wheels when climbing = HG' W12 Exploring wheel Obstacle G1 R0 W11 OpenWHEEL OW i3r Stability R1 G W22 R2 P21 P12 H G' G2 W21 P22 15

16 process sequence in 7 phases / 19 stages 1 A - Prepairing W22 Low 2 W12 B - W 11 climbing C - W 12 climbing E - W 21 climbing High W21 W11 6 OpenWHEEL OW i3r Stability 10 D Going forward 15 F - W 22 climbing G - 16

17 process Strong simplifying hypotheses Negligible mass of the inter-axle mechanism Non-deformable bodies (i.e. infinite part stiffness) Small warping rotation-angles to avoid representation of complex 3D poses Punctual ground-wheel contact with toric wheels Perfect rolling without slipping assuming that normal forces are sufficient to ensure enough traction OpenWHEEL OW i3r Stability Incremental validation Multi-body Adams model [IROS 06] Reduced model [MTM 2008] Full scale model [in process...] 17

18 Wheel sub-assembly Four identical wheel sub-assemblies One 9V actuator : 200 rpm, 3.52 N.cm on top of the wheel Transmission ratio 1/15 Rubber air-tire with good friction (Diam 49.6 mm) Overconstrained structure - mass 149 g 2 Z40 Z8 3 C Wheel Inter-axle Whole robot E 0 Z8 Software Z

19 Inter-axle mechanism High torque warping mechanism The same 9V actuator as for the wheels Transmission ratio 1/560 Double worm gear redundant overconstrained transmission Minimized backlash Improved tooth strength Z16 Wheel Inter-axle Whole robot F1 Z16 Z8 E F2 Z8 Software Z56 Z40 19

20 Whole reduced model of the robot Whole assembly = 2 axles + 1 inter-axle mechanism Total weight 1430 g Center of mass quasicentered G1 G=0.497 G1 G2 Wheel Inter-axle Whole robot Software Translation speed 55mm/s Warping speed 45 in 21s (with oil) 170 mm G1 Carries its power (12 AA batteries) G G2 W m m W m m 20

21 Software architecture Embedded program in each control unit NQC language, BricxCC developing environment unit 1 for axle 1 (Master) unit 2 for axle 2 and warping joint R0 (Slave) Exchanges between units via infrared port Protocol by message sending and detection loop Master program for A1 Wheel Inter-axle Whole robot // Stage 2 : rev W22 / fwd W21 Bip(); SendMessage (2); ClearMessage(); until (Message() == 2); Slave program for A2+ R0 // Stage 2 : rev W22 / fwd W21 if (Message() == 2) { OnRev (W22); OnFwd (W21); Wait(150); Off (W22+W21); SendMessage(2); } Software 21

22 al climbing Purpose: validating the climbing strategy Obstacle: 55 mm high Higher than a wheel Actuators at full speed with open loop control Useful to determine the most suitable sensors Adjustments Improvement Difficulties Initial tests with stabilized power Final test with batteries Difficult to debug: the final pose depends on the full process 22

23 Phase A Stage 01 Adjustments Improvement 23

24 Phase B Stage 02 Adjustments Improvement 24

25 Phase B Stage 03 Adjustments Optional: going against the obstacle and actuating the wheel for bonus tangential force Improvement 25

26 Phase B Stage 04 Adjustments Improvement 26

27 Phase B Stage 05 Adjustments Optional: the exploring wheel can land faster if it was lifted just at the level of the obstacle Improvement 27

28 Phase C Stage 06 Adjustments With pitch angle and contact on four wheels, steering Axle A1 is coupled with warping the frame Improvement 28

29 Phase C Stage 07 Adjustments W11 serves as a pivot for Axle A1 Slipping risk on W11 Improvement Solution: unsteer slightly Axle A2 Slightly lifts W12 Increase normal force on W11 29

30 Phase C Stage 08 Adjustments Improvement 30

31 Phase C Stage 09 Adjustments Improvement 31

32 Phase D Stage 10 4W Stable W and W with same speed set point Normal forces differ Induced steering on axle A1 Adjustments Improvement Solution: closed-loop control - to equilibrate normal forces - to keep a constant steering angle 32

33 Phase E Stage 11 Adjustments Improvement 33

34 Phase E Stage 12 Solution: additional 149 g counterweight Risk of instability during W21 climbing Adjustments Improvement 34

35 Phase E Stage 13 Adjustments Improvement 35

36 Phase E Stage 14 Optional: the exploring wheel can land faster if it was lifted just at the Adjustments level of the obstacle Improvement 36

37 Phase F Stage 15 Optional: going against the Adjustments obstacle and actuating the wheel for bonus tangential force Improvement 37

38 Phase F Stage 16 Riskrobots of instability Agile during W22 climbing Solution: additional 149 g counterweight Adjustments Improvement 38

39 Phase F Stage 17 Adjustments Improvement 39

40 Phase F Stage 18 Adjustments Improvement 40

41 Phase G Stage 19 Adjustments Risk of lateral drift with respect to phase 1 Improvement 41

42 Design adjustments Axle A1 climbs more easily than Axle A2 Need for a counterweight The counterweight breaks longitudinal symmetry Possible explanation: if the centre of mass G is too high, its projection on the ground G' moves relatively to contact points Pi and stability criterion is no more respected G2 G G1 G G2 Adjustments Improvement G1 P1 P2 G' P1 P2 G' The 149 g counterweight brings G forward of 16 mm G1 G=m2 / m1 m2 G 1 G2=0.408 G1 G2 42

43 laws for the 4 wheels and central actuator Phases B-D-E-F are similar in length Warping phases take 80% of time. Warping angle < Phase E Angles of rotation of the actuators of OW i3r 600 Angle Angle Angle Angle Angle 500 Phase F W12 ( ) W11 ( ) W22 ( ) W21 ( ) R0 ( ) 400 Phase B Phase C Adjustments Improvement s 43

44 improvement The good metrics for measuring climbing ability Better control to improve climbing Adding sensors for precise monitoring Adjustments Not the wheel diameter Comparison Obstacle height / Altitude of the centre of mass OpenWHEEL i3r can climb obstacles as high as 67% ZG Angular coders on actuators (wheels + warping central joint) Coders on passive joints (axle steering) Ultra-sound sensor to detect obstacle / measure height Rolling without slipping Measuring normal force Pitch Two-axes force gauges in the rim of the wheels Improvement 44

45 Results A new principle for stable obstacle climbing Usable with only 4 wheels for simplicity Only one supplemental central actuator A climbing process in 19 stages Validated on a reduced model of OpenWHEEL i3r Climbs obstacles as high as 66% of ZCentre of mass In the future Geometrical model for coupled actuation of steering / warping Obstacle detection and control adaptation Normal force / slipping regulation Optimizing kinematics & structure Optimizing climbing strategy: how much can we climb with 4 wheels? 45

46 Next step? 06 A stable climbing process Multibody validation (Adams) al validation on actuated reduced model 46

47 Full scale demonstrator 140 kg, 1.4m long 47