www.servomagazine.com/index.php?/magazine/article/august2011_GeerHead
GEERHEAD
Laboratories paper by Tom Moore,
et. al.
Hardware
Configuration,
Environmental
Mapping
Covert robot
rear angle.
The Lockheed Martin engineers
added a motherboard with a 2. 4 GHz
dual core chip, a power supply, and a
hard drive to the Pioneer robot. Using
an audio capture card, the robot can
detect and process sounds which
point to the position and approach of
any sentries at the target location. The
microphones are mounted at the
highest possible location within the
robot’s dimensions. The engineers
used a mic preamp to boost signals to
the sound card.
A long range LIDAR is mounted
even higher up on the robot, while a
short range LIDAR was mounted
under the robot. These detect distant
obstacles and immediate obstacles,
respectively. A 6-DOF, gyro stabilized GPS/IMU unit
mounted atop the robot provides the course setting,
attitude, and vehicle speed data.
The robot uses “cost maps” on a 2D grid that hold the
locations of obstacles. The low obstacles are identified as
places the robot could not hide, while the high obstacles
are noted as places the robot could hide from a sentry.
There are different types of cost maps too, including
individual intervisibility, generalized intervisibility, light levels,
desirability, and one that shows how easy it would be for
the robot to run away if a sentry was nearing it
(escapability).
The desirability map demonstrates the degree to which
the target location to be viewed is visible from a given cell.
Before any mission, the operator draws a circle around the
target area that the robot should get a good view of. The
cells that are part of the target area are initially considered
to be desirable.
While the desirability map helps the robot to choose
goals that are hidden yet allow line-of-sight with the target
location, it does not consider locations that enable the
robot to easily escape. It does not identify, for example, a
dead end in an alley as inescapable. The escapability map
shows how easily the robot could escape from any given
cell. Using both maps, the robot selects goals or cells that
are both desirable and escapable, narrowing its options to
the best overall. Preferably, the robot will have several
directions by which it can escape a guard.
Escapability is determined by cells where opposing and
orthogonal rays of substantial length pass through.
Escapability calculations and generalized intervisibility
calculations (discussed later) have an inverse relationship.
While the cell or area that is most escapable has rays of
infinite length in all directions (not stopped by obstacles),
these areas or cells would be the worst for intervisibility.
Combining Map Data To
Achieve A Target Goal
Based on the cost maps, the robot determines the
overall stealth cost, escapability, and desirability of each cell,
and calculates its goals accordingly. Obviously, the robot
selects the cells with the least stealth cost and the greatest
desirability and escapability over other cells when planning
its goals and course of movement.
Since the robot is designed to take advantage of low
light/no light in nighttime operations, it considers light
sources when navigating and avoids them as much as
possible, as well. The robot seeks out the cover of shadows
and avoids light sources while maintaining considerations
for the various maps.
The robot uses a map of existing light sources in a
region rather than detecting light sources in real time. This
makes flashlights, headlights, and other unpredictable
movable light sources a challenge. Light sources are
assigned costs based on the cell through which the light
passes. Each additional light source per cell adds to the cost
of a cell with only one light source.
SERVO 08.2011 11