The Wayfarer Modular Navigation Payload for Intelligent Robot Infrastructure

1
The Wayfarer Modular Navigation Payload for
Intelligent Robot Infrastructure

• Brian Yamauchi
• yamauchi_at_irobot.com

2
Autonomous Urban Recon

• Critical Army need
• Daily casualties during patrols
• Hazards
• Snipers
• Improvised explosive devices
• RPGs
• Teleop is not enough
• Limited comms range
• Line-of-sight restrictions
• Limited bandwidth

3
Modular Navigation Infrastructure
PackBot ?
FCS SUGV ?
Wayfarer Modular Navigation Payload
? R-Gator
4
Wayfarer Objectives

• Focused, applied research project
• Develop technology to enable small UGVs to
  perform autonomous recon missions
• Enable small UGVs to
• Navigate autonomously down urban streets
• Record digital video (EO/IR) and build map
• Return autonomously to starting point
• Provide video log and map to warfighters

5
Urban Reconnaissance Missions

• Perimeter Recon
• Follow outside wall of building
• Route Recon
• Follow current street for specified distance
• Street Recon
• Follow GPS waypoints and street directions
• Record EO/IR video and build map
• Return to starting point

6
Wayfarer Sensors

• SICK LD OEM 360-degree planar LADAR
• Point Grey Bumblebee stereo vision system
• Organic GPS receiver
• Crossbow six-axis IMU
• Color and low-light B/W video cameras
• Indigo Omega FLIR camera
• Swiss Ranger 3D flash LADAR

7
Wayfarer PackBot
8
Scaled Vector Field Histogram (SVFH)

• Extension of Borenstein Korens Vector Field
  Histogram (VFH) obstacle avoidance
• VFH Map each range reading from sensors to
  corresponding polar coordinate sector
• SVFH Spread each reading to vote for all sectors
  within an arc length inversely proportional to
  range
• q k / r
• q arc length (radians),
• k constant (0.4), r range (meters)

9
SVFH Hallway Example
SVFH Bins
Clear Vectors
10
SVFH Intersection Example
SVFH Bins
Clear Vectors
11
Wayfarer Obstacle Avoidance
12
Perimeter Following

• Detect and follow building walls
• Hough Transform
• Detects linear features in range data
• Finds building walls and other street-aligned
  features (e.g. cars, curbs)
• Collects votes to determine current wall heading
• Integrated with obstacle avoidance

13
Hallway Wall Detection
Best Line
All Lines
14
Intersection Wall Detection
Best Line
All Lines
15
Outdoor Perimeter Detection
Wall Tracking
Landscape Tracking
16
Wayfarer Perimeter Following
17
Autonomous Mapping

• Build occupancy grid map on robot
• Transmit local region back to OCU to minimize
  bandwidth usage
• Runs in parallel with navigation and avoidance
• Robot can travel beyond communications range and
  return with a map
• Can also run in background during teleoperation

18
Wayfarer Autonomous Mapping
19
Indoor Map
20
Localization

• Heading tracking using Hough transform
• Track orientation of linear features
• Use to estimate robot orientation
• Alternative approaches
• Scan matching
• GPS/Compass/INS
• Kalman filters
• Particle filters (Monte Carlo SLAM)

21
Future Work

• Outdoor testing of perimeter following
• Street following
• GPS-based street navigation

22
Conclusions

• Wayfarer will provide autonomous urban navigation
  technology for man-portable UGVs
• By September 2005, we will have two
  fully-operational Wayfarer UGV prototypes able to
  perform urban reconnaissance missions
• Wayfarer navigation payload will provide
  intelligent navigation infrastructure for
  PackBots and R-Gators, as well as FCS SUGV and
  other UGVs