Guidance of Autonomous Tractor With Four Wheel Steering Tim o - PowerPoint PPT Presentation

Guidance of Autonomous Tractor With Four Wheel Steering Tim o Oksanen Doctor of Science (technology) ( gy) Docent (agricultural engineering) Senior Research Scientist Senior Research Scientist Aalto University, Finland Dept of Automation and Systems Technology Dept. of Automation and Systems Technology IEEE RAS AgRobots TC IEEE RAS AgRobots TC Webinar #11 September 26, 2013

The presentation The presentation • Most of the slides related to presentations in: p – Oksanen, T., Backman, J. 2013. Guidance system for agricultural tractor with four wheel steering. IFAC Bio-Robotics Conference, Sakai Japan 27 29 March 2013 Sakai, Japan, 27-29 March 2013. – Oksanen, T. 2012. Embedded control system for large scale unmanned tractor. 5th Automation Technology for Off-road Equipment Conference (ATOE), Valencia, Spain, July 8 - July 12, 2012. pp. 3-8. – Oksanen, T. 2012. Path following algorithm for four wheel independent steered tractor 5th Automation Technology for Off-road independent steered tractor. 5th Automation Technology for Off-road Equipment Conference (ATOE), Valencia, Spain, July 8 - July 12, 2012. pp. 9-14. IEEE RAS TC on Agricultural Robotics and Automation Webinar #011 Timo Oksanen 26.9.2013 slide 2

Part 1: The tractor "APU Module" Part 1: The tractor APU-Module slide 3

The Autonomous Tractor 165 hp turbodiesel ~6000 kg g "APU Module" APU-Module Hydraulic drivetrain Each wheel ( 4WD ) • Steering Steering • Drive 3p-hitch in both ends PTO in both ends PTO in both ends Up to 9 aux valves 12V DC electric Built originally 1990-1992 by a Finnish company. E&E refurbished 2011-2012 2011 2012 Originally designed for unmanned use unmanned use (autonomous) IEEE RAS TC on Agricultural Robotics and Automation Webinar #011 Timo Oksanen 26.9.2013 slide 4

Drivetrain Drivetrain • Four variable displacement hydraulic pumps attached to shaft of diesel engine • Constant displacement motors in each wheel hub • One pump drives one motor, independent control of wheel drive (coupled only by ground contact) – ”differential” needs to be realized in electronic control system • Encoder in each wheel to measure speed • Each wheel has independent hydraulic actuator for steering with position sensor; no track rods used – makes it possible to achieve accurate Ackermann geometry in k it ibl t hi t A k t i four wheel steering • � 4 wheel drive + 4 wheel steering � 4 wheel drive + 4 wheel steering IEEE RAS TC on Agricultural Robotics and Automation Webinar #011 Timo Oksanen 26.9.2013 slide 5

Drivetrain Drivetrain • Original design of tractor did not contain brakes at all g g – Hydrostatic system is able to do deceleration – When the tractor is stationary, a small drift happens � Parking brake implemented in rear wheels • For implements the original hydraulic system provides p g y y p 180 l/min @ 200 bar hydraulic flow – up to 9 auxiliary hydraulic valves available in both ends – analogue proportional heads in the directional valves • The diesel engine is from 1990 (Perkins 1006-6T) g ( ) – No built-in ECU – Monitoring and control of engine needs to be realized IEEE RAS TC on Agricultural Robotics and Automation Webinar #011 Timo Oksanen 26.9.2013 slide 6

Requirements for Electronic control system Requirements for Electronic control system • System y – Real-timeness – Safety • Control interfaces – Interface to autonomous navigation system g y – Wireless manual control (safety) • Functions – Cruise control – Coupled steering & drive p g – Brake control – Engine control – Hitch and PTO control IEEE RAS TC on Agricultural Robotics and Automation Webinar #011 Timo Oksanen 26.9.2013 slide 7

Materials Materials • The tractor has plenty of I/O; altogether ~80, of which 34 need to be PWM (>2A) for hydraulic propos d t b PWM ( 2A) f h d li • A control module Parker/Mitron MCC2212 was selected due to several reasons – Plenty of power outputs (12 DO + 10 PWM) – Programming with C language – Pretty mature product, more than 10 years – (the microprocessor itself is outdated) (th i it lf i td t d) – CAN bus • ≥ 4 of these are needed (I/O) ≥ 4 of these are needed (I/O) • Communication – adapted from SAE J1939, mainly 100ms IEEE RAS TC on Agricultural Robotics and Automation Webinar #011 Timo Oksanen 26.9.2013 slide 8

Control system Control system IEEE RAS TC on Agricultural Robotics and Automation Webinar #011 Timo Oksanen 26.9.2013 slide 9

Part 2: Guidance & four wheel steering Part 2: Guidance & four wheel steering slide 11

Introduction Introduction • The presentation shows how to keep a vehicle with 4WS p p on track on the field • Later the results in real field operation Later the results in real field operation • Four wheel steering is used in some commercial tractors: tractors: Case 4894 Claas Xerion series IEEE RAS TC on Agricultural Robotics and Automation Webinar #011 Timo Oksanen 26.9.2013 slide 12

Seed drill (combined) Seed drill (combined) Tume KL-2500 (1987) • 125 mm seed coulters • 250 mm fertilizer coulters • 0.46 m 3 fertilizer • 0.31 m 3 seeds • Cat2 hitch mounting IEEE RAS TC on Agricultural Robotics and Automation Webinar #011 Timo Oksanen 26.9.2013 slide 13

Materials Materials • Constraints – Steering angle (~20 deg) – Steering rate (8-12 °/s @1500rpm) – Steering dynamics (delay ~400ms) St i d i (d l 400 ) • � limits the driving speed vs. tracking accuracy – Vehicle max. speed cannot be utilized (40 km/h) V hi l d b ili d (40 k /h) • Positioning devices a.k.a. "HighDock" – RTK-GPS (VRS), Trimble 5700 – Fiber-optic gyroscope, KVH DSP-3000 (for heading) – Inclinometer, Inertial-Link 3DM-GX2 (incl. MEMS gyros) I li t I ti l Li k 3DM GX2 (i l MEMS ) – Low-level sensor fusion in heading estimation • Communication: 2x 250kbps CAN-bus, 100 ms cycle time C i ti 2 250kb CAN b 100 l ti IEEE RAS TC on Agricultural Robotics and Automation Webinar #011 Timo Oksanen 26.9.2013 slide 14

Path tracking Path tracking • Route planner gives waypoints five seconds ahead p g yp – Polyline (x, y), speed, acceleration limit, working position, heading offset – The polyline is feasible (e.g. turning radius) • Error variables in tracking – Lateral error – Angular error • While navigating, the waypoints are removed as soon as they are passed IEEE RAS TC on Agricultural Robotics and Automation Webinar #011 Timo Oksanen 26.9.2013 slide 15

Prediction Prediction • The state of robot predicted over finite time horizon • Constant velocity and zero steering assumed – Through dynamic model • Path error calculated for each predicted state p � vector of errors • Weighted average Weighted average to form errors for navigation o a ga o IEEE RAS TC on Agricultural Robotics and Automation Webinar #011 Timo Oksanen 26.9.2013 slide 16

Structure of inverse kinematic path tracker Structure of inverse kinematic path tracker Path Error Path Error x x v Lateral Calculation α F Controller Inverse α R Kinematics Angular Approach Controller Filter Feed- forward forward IEEE RAS TC on Agricultural Robotics and Automation Webinar #011 Timo Oksanen 26.9.2013 slide 17

The Approach Filter The Approach Filter • The regulators are minimizing angular error and lateral g g g error, separately • In case the lateral error is (very) large, ”using” only the In case the lateral error is (very) large, using only the crab steering manner for getting on the track would take a long way (as the steering angle is limited) g y ( g g ) • When lateral error is large, the Approach Filter modifies angular error signal angular error signal IEEE RAS TC on Agricultural Robotics and Automation Webinar #011 Timo Oksanen 26.9.2013 slide 18

Part 3: Field experiments Part 3: Field experiments slide 19

Field experiment plan (autumn 2012) Field experiment plan (autumn 2012) • Preplanned route p • 2.4 ha winter wheat • Strategy • Strategy 1. six times around the field; CCW, the field; CCW, reversing in corners 2. looping the rest, skipping 5-6 swaths • Refilling tanks manually • Latitude 60.45° IEEE RAS TC on Agricultural Robotics and Automation Webinar #011 Timo Oksanen 26.9.2013 slide 20

the path is linear piecewise (not smooth) Accuracy Accuracy 5 5 0 gle (deg) -5 ang -1 0 1 2 0 0 1 2 0 0 1 2 5 0 1 2 5 0 1 3 0 0 1 3 0 0 1 3 5 0 1 3 5 0 1 4 0 0 1 4 0 0 0 . 5 0 teral (m) lat -0 . 5 -1 1 2 0 0 1 2 0 0 1 2 5 0 1 2 5 0 1 3 0 0 1 3 0 0 1 3 5 0 1 3 5 0 1 4 0 0 1 4 0 0 t im e (s ) IEEE RAS TC on Agricultural Robotics and Automation Webinar #011 Timo Oksanen 26.9.2013 slide 21

Accuracy Accuracy 1 . 5 1 gle (deg) 0 . 5 0 ang -0 . 5 -1 -1 . 5 1 2 0 0 1 2 0 0 1 2 5 0 1 2 5 0 1 3 0 0 1 3 0 0 1 3 5 0 1 3 5 0 1 4 0 0 1 4 0 0 0 . 0 5 eral (m) 0 late -0 . 0 5 1 2 0 0 1 2 0 0 1 2 5 0 1 2 5 0 1 3 0 0 1 3 0 0 1 3 5 0 1 3 5 0 1 4 0 0 1 4 0 0 t im e (s ) IEEE RAS TC on Agricultural Robotics and Automation Webinar #011 Timo Oksanen 26.9.2013 slide 22