Reliable Operations of Unmanned Ground Vehicles: … Robotics Reliability Center University of Michigan College of Engineering GRRC # 1 !! Reliable Operations of ! Unmanned Ground

Ground Robotics Reliability Center University of Michigan College of Engineering

GRRC # 1

!

!Reliable Operations of !

Unmanned Ground Vehicles: !Research at the !

Ground Robotics Reliability Center!!

http://grrc.engin.umich.edu/ !!

Dawn M. Tilbury, Professor!Mechanical Engineering!

College of Engineering, University of Michigan, Ann Arbor!!!

College of Engineering !University of Michigan!Ann Arbor!

TARDEC: U.S. Army Tank Automotive Research, Development and Engineering Center


GRRC # 2

Reliable Operations!

Critical Unmet Needs in Unmanned Ground Vehicles and Robotics!•  Tactical Behaviors •  Autonomous Controls •  360 Degree Awareness •  Safe Operations of Robots Around Humans •  Reliability & Manufacturing •  Machine-Human Interface •  Platform Mobility •  Tele-Operation

Milestones of the GRRC •  Established with a few projects in 2007 • Kick-off as GRRC in August 2008 •  iRobot as industry member; PackBot donation •  Focus on Reliable Operations of Ground Robotics Center Goals: • Develop research to support reliable performance and operation of

unmanned ground vehicles (UGVs) and robots. •  Educate graduate students in reliable design and operation of

UGVs through research projects, curriculum and courses. •  Increase the impact and speed of delivering cutting-edge reliable

UGV technology readiness for the soldier

Ground Robotics!Reliability Center


GRRC # 4

Challenges in UGV Reliable Operation!

•  Robotics has grown out of a hobbyist mindset (i.e., prototypes, demos)!

•  Serious reliability problems!–  Low mean-time-between-failures (reported

times of 6-24 hrs)!•  Lack of existing research in reliable

operations of UGV!–  Automotive industry produces mature,

complex products!–  Industrial manipulators perform repetitive

tasks in structured environments!•  Goal: Develop research to improve

reliable operation of UGVs!!

“The robotics industry faces many of the same challenges that the personal computer business faced 30 years ago”


GRRC # 5

Four Pillars of Reliable Operations!New types of sensors not proven for usage in the field

Manufacturing defects cause component failures (pinched wires in assembly)

Obstacles Slopes, ditches Dust, mud, rocks Temperature

Drive up steep slope Turn left instead of right Exceed motor limits

Manufacturing Design

User Environment

Reliable UGV

Operations

Specification

Technology

Standards Procedures

Human

Interference

Hazard

Impact

Mistakes

Slips

Machines

Interface


GRRC # 6

Reliable Behavior by Instruction!•  Interactive instruction!

–  Human guides robot during mission!

–  Robot can independently perform known tasks!

–  Robot requests guidance when unfamiliar situation occurs!

•  Situational learning !–  Leverage context and

shared knowledge!

Laird


GRRC # 7

Adjustable Autonomy!•  Cooperative control between robot and

operator!–  Robot has autonomous behaviors!–  Can ask questions of the operator to gain

information about environment!•  Markov Decision

Process!–  Reward for states!–  Cost for questions!–  Goal: maximize

expected reward !

UGV

Des(na(on

Human Crowd

Human Crowd

UAV

Durfee and Bajeva


GRRC # 8

Exploration and Inventory of Human-Filled Spaced!•  Long-term goals:!

–  Robots in an unknown environment!

–  Build a map, identify people, objects!

–  Coordinate with commanders!

•  Current work: Sensor fusion!–  Laser plus vision!

Olson


GRRC # 9

Exploration and Inventory of Human-Filled Spaces!

2D segments extracted using image-only method: Some segment boundaries missing.

Segmentation using only a color difference criterion: Failure to segment floor and trash bin correctly.

Segmentation with only a surface normal criterion: Rear wall has missing segments

Our Method, which combines both laser and camera data.

Olson


GRRC # 10

•  Team Michigan with 14 robots, competition in Australia!

•  Autonomous localization,map-building, OOI detection !

•  Minimal humanintervention!

MAGIC Winners!Olson


GRRC # 11

Augmented/Virtual Reality Interfaces!•  Improve operator performance and reliability

through improved user interface!–  Add augmented reality to video feed to indicate

distance to obstacles!–  Add virtual reality for a “birds-eye” view of robot

in workspace!

Augmented Reality Scene Virtual Reality Scene Robot Platform

Tilbury


GRRC # 12

Power/energy for mobile robots!Peng & Filipi


GRRC # 13

Integrated Power Systems for Improved Mobility!•  Model loads on mobile robots!

–  Sensors, communication, processors, etc.!–  Off-road driving cycles characterization!

•  Model energy conversion/storage devices!–  Understand limitations!–  Design guidelines!

•  Leverage hybrid cardesign methods!

•  Goal: 8-hour missionduration for UGV!

Peng & Filipi


GRRC # 14

•  Goal: Inspect an area with autonomous UGV while minimizing energy usage!

•  Approach:!–  Compare energy usage along

existing coverage paths!–  Adapt paths to further reduce

energy needs!•  Gradual turns instead of

turn-in-place!•  Optimize velocity profile along

straight-line segments!

Energy-Aware Coverage Control!Atkins & Tilbury

Boustrophedon path

Spanning Tree path


GRRC # 15

UGV testing for reliability !

•  Current: Pass/fail for robot capabilities under pre-designed conditions!

•  Our approach: Simulation-based “testing” to determine robot operating limits w/o failure!

o  NIST Test Arena o  Aberdeen Test Center

Jin & Ulsoy


GRRC # 16

Simulation-based testing!•  Simplified model of Packbot,

dynamic model in ADAMS!•  Determine rollover limit with

different manipulator positions!

Jin & Ulsoy


GRRC # 17

Robust mechanical design!•  Adding a torsional spring can reduce the

required torque on the manipulator arm!•  Design optimized for nominal trajectory!•  Q: How does the extra complexity affect

reliability?!

Figure 2) Model joint drive train schematic.

New design

New model

Ulsoy


GRRC # 18

Moving Obstacle Avoidance with Sensor Uncertainty!•  Velocity occupancy space gives weights

based on probability of collision with obstacles for different robot velocities!

•  Laser-scan data to estimate obstacles position and velocities!

•  Combine negative weights of obstacles with positive weightfor goal position!

Peng & Ulsoy


GRRC # 19

Indoor position tracking!•  GPS unavailable, video feed disorienting!•  Solution: Dead-reckoning plus heuristics!

(a)

(b)

Figure 1: With our newly proposed Heuristics-enhanced Odometry (HEO) method, operators of remotely-controlled mobile robots will see the trajectory of their robot on their console, even in GPS-denied, indoor environments. (a) Conventional odometry-derived trajectory in a GPS denied indoor environment. (b) Trajectory of same indoor trip, but now with HEO.

Same gyro & odometry, after applying our

heuristics algorithm.

(a)

(b)

Figure 1: With our newly proposed Heuristics-enhanced Odometry (HEO) method, operators of remotely-controlled mobile robots will see the trajectory of their robot on their console, even in GPS-denied, indoor environments. (a) Conventional odometry-derived trajectory in a GPS denied indoor environment. (b) Trajectory of same indoor trip, but now with HEO.

Conventional gyro & odometry-derived trajectory in a GPS-denied

indoor environment.

Borenstein


GRRC # 20

Leader-Follower Navigation without Line-of-Sight!•  Personal dead-reckoning on leader plus

simple IMU on robot follower!•  To avoid unbounded heading errors, impose

synchronization on heading!•  Up to one minute lag accommodated!

Borenstein


GRRC # 21

Robotics testbeds!•  Packbot with AWARE2

software environment!•  SuperDroid wheeled and

tracked robots!!

UM’s PackBot (donated by iRobot)

SuperDroid wheeled ATR robot

kit. Payload: >20 lbs

SuperDroid LT tracked ATR robot kit. Payload >10 lbs

Borenstein


GRRC # 22

Robotics testbeds!•  Integrated

power testbed !

Pan/Tilt/Zoom Video Camera

3D Laser Range Finder(HOKUYO UTM-30-LX)

Solar Panel (expandable)

200 × 50 mmAll-terrain

Pneumatic tire

Double WishboneIndependent Suspension

On-board Computer and

Electronics

Li-ion Battery Pack

Fuel Tank

Muffler Modified Model Engine

Starter/Alternator

Power Electronics

Maxon EC Motor(4 Quadrant Operation, Regenerative Braking)

AdjustableShock Absorber

• Solar Integrated Plug-in Hybrid Power System• Extended mission time and capability; Quick refueling• Battery charged on board or recharge dock• Autonomous energy-sustaining strategy• Silent mode; Regenerative Braking

• Improved Mobility• Skid-steering• 6 Wheel Drive• Independent Suspension• Travel over Rough terrain• Stair Climbing

Peng & Filipi


GRRC # 23

Summary!•  Research in Reliable Operations of UGVs!•  Four research areas!

–  Behavior reliability!–  Power and mobility!–  Design and control for reliability!–  Testbeds, integration, and validation!

•  Partner with US-Army TARDEC and Industry!

Documents

Reliable Operations of Unmanned Ground Vehicles: … Robotics Reliability Center University of Michigan College of Engineering GRRC # 1 !! Reliable Operations of ! Unmanned Ground