Download - Intelligent off-road vehicles Martin Servin, Department of Physics, 2008-04-02 \proj\ifor \proj\ifor
Intelligent off-road vehiclesMartin Servin, Department of Physics, 2008-04-02
www.umu.se\proj\ifor
Outline
• Background to the field• Overview IFOR• Autonomous navigation• Crane automation• Simulator based design
Feel free to ask questions and make comments and proposals!
A sample of technological gems…
Mars rover – extreme teleoperation
Deep Blue – reasoning computer
DARPA Grand Challenge – competition with autonomous vehicles
QRIO – balancing robot
Parthenon – virtual 3D reconstruction
HCI – retinal display
The off-road challangeDemand for new technology• Increased productivity• Increased safety and work environment• Environmental sustainability
The forestry challange• Complex work processes to automate
– no computer beats the human in running a harvester• Rough environment with big variations
– sensor vision in forest, robust and sustainable system handling vibrations, moist and dirt
Vision from forestry industry
“2025 – Ingen man i maskinen, ingen hand på spakarna “
• An initiavive for R&D for intelligent off-road technology
• Initiated by the industry in 2001
• Collaboration between academia and industry
= a forum for R&D and a collection of projects focused at IFOR technology
What is IFOR ?
Academia:Umeå University
Swedish University of Agricultural Sciences
Skogforsk
Industry:Komatsu Forest
Holmen Skog
Sveaskog
BAE Systems Hägglunds
LKAB
+ network of other research centers and companies
Technology vision
• Improved work environment
• Increased productivity and cut costs
• Increased safety
• Reduced environmental impact
Technology:Technology:Goals:Goals:
• Control algorithms and modeling
• Interaction – man, machine and environment
• Sensor vision
• Localization and map building
20012001 2025202520120100
Automation Automation of routine of routine work work processesprocesses
Crane tip Crane tip controlcontrol
UnmanneUnmanned vehicles d vehicles
Activities and projects
Autonomous navigationDr Thomas Hellström1 PhD studentsComputing Science Department
Smart Crane ControlProf Anton Shiriaev1 FoAss, 1 PostDoc, 3 PhDDepartment of Applied Physics and Electronics
Vehicle simulatorsDr Martin ServinIn collaboration with VRlab at UmU
Miscellanious- Seminars and workshops- Experiments and pre-studies- Student projects
Equipment• Forest machines – valmet forwarder and harvester• Full sized in-door hydraulic crane• Portable prototyping hardware for feedback control • Sensors (dgps, laser radar, hydraulic pressure,
stereo camera,…) • Simulator systems
Funding> 25 MSEK since 2001Kempe Foundation, Sveaskog, Vinnova,Komatsu Forest, Sparbanksstiftelsen Norrland, Umeå University, LKAB, BAE Systems Hägglunds
Other actors
SLU
Skogforsk
Applied Mathematics – Prof Mats G Larsson
Design Institute
UCIT / ProcessIT Innovations
Autonomous navigation
Autonomous navigationDr Thomas Hellström
1 PhD students
- unmanned transportation of logs
- localization, path tracking and path planning
- RTK-DPGS with cm accuracy
- laser scanners, radars,...
- machine learning and sensor fusion
- first prototype demonstrated in Dec 2005
- ”Simulator in the loop”
Autumn 2008 we are running the student DBT-projects:
- Sensor vision and remote operation
- Simulation of terrain vehicle with autonomous abilities
www.cs.umu.se/research/ifor/IFORnav/videos.htmwww.cs.umu.se/research/ifor/IFORnav/videos.htm
Smart Crane Control
Smart Crane ControlProf Anton Shiriaev – Control System Theory1 FoAss, 1 PostDoc, 3 PhD
- motion planning, motion control for mechanical systems- feedback design for hydraulically actuated cranes- crane tip control- optimized motions – speed and stability- semi-automation, e.g. automatich loading- VR-enabled remote operation- portable prototyping hardware for feedback control
Recent results:- motion faster and more stable than human operator – Valmet forwarder- demonstrated automatic loading in lab
Grant from “Stiftelsen för strategisk forskning” for crane control using only hydraulic measurements at Komatsu Forest
1 industrial PhD have been granted (?) - Komatsu Forest and Umeå University splitting the costs 50-50 – Semi-autonomous harvester control system
Visual Simulation of Machine Concepts for Forest Biomass Harvesting
Martin Servin, A. Backman, K. Bodin - Umeå University, SwedenU. Bergsten, D. Bergström, T. Nordfjell, I. Wästerlund - Swedish University of Agricultural Sciences, Sweden B. Löfgren - Skogforsk (the Forestry Research Institute of Sweden)
VRIC 2008 – 10th International Conference on Virtual Reality (Laval Virtual)
Outline• Simulator-based design• Forest biomass harvesting
– concept machine and work method
• Experiments in simulator environment– system and procedure– purpose: find optimal harvesting technique and machine design
Training simulator technology – also for concieving new machines concepts and work methods
Simulator-based design (SBD)
Simulation tools are converging – R&D process impoves – cross-disciplinary participation
• Extension of virtual prototyping and simulation to include human-in-the-loop• Fast and sheap• Simulators – complex yet controllable environments
Figure from T. Alm ”Simulator-based design” (2007).
End customer Manufacturer Designer ResearcherEngineer
Simulator training
Application of SBD to:
Forest biomass harvesting
• Increasing demand for forest biomass• Early harvesting/thinning is becoming profitable• Large volumes and areas, small income per unit, energy consumption• Crucial to use optimized technology – economically and environmentally sustainable
• Uncertain on what solution to choose for thinning
• Virtual and real prototypes are important!
New harvesting methods in dense forest stands
Early harvesting = thinning + biomass harvesting - single-tree harvesting - multi-tree harvesting - geometric area based felling
strip roads 3 m wide every 15-20 mcorridors 1x10 m10 trees, 6 m, 50 kgcollect in piles of 50 trees
Machine concept for harvesting in dense forest stands
Size: 4x2 m, 2.5 ton, 8m reachMobility: indv 4W on pendulum armsHarvester head: multi-tree vs bladeControl and HMI: boom-tip control, semi-autonomous,teleoperation (direct or VE), laser scanner & stereo camera, dynamic 3D maps from sattelite and AUV
Experiments in simulator environment- system and procedure
System componentssoftware: Colosseum3D (OSG, Vortex – AgX Multiphysics, lua,…)hardware: full simulator environment (screen projection,
authentic chair and joysticks, motion platform) or portable case, convential multicore PC
models: data from real forest stands in 3D terrain, vehicle = 20 rigid bodies coupled by kinemtaic constraints (wheel suspension, crane joints,…)
vehicle automation and HMI module: vehicle control, automation, sensor, 3D-map engine and HMI interface
The application requires advanced real-time physics: terramechanics, stacking, hydraulics,…
Experiments in simulator environment- system and procedure
Experiment procedureTask: do harvest thinning in a given dense forest stand
Variations:- forest stand (distribution, species, topology)- harvester head (single, multi, sword)- vehicle (existing machines, new proposals)- automation and HMI (manual, semi-automatic, fully auto)- operator
Measurements: - time per biomass unit in kg (strip road, corridor, tree, move to pile, positioning, transport)- energy consumption- work environment
Optimize: find optimal mechine design and work method – data from simulator experiments used in logistics computation