> « back to results for ""
Below is a cache of http://www.darpa.mil/grandchallenge04/teamtechpapers/teamcyberriderfinaltp.pdf. It's a snapshot of the page taken as our search engine crawled the Web.
The web site itself may have changed. You can check the current page or check for previous versions at the Internet Archive. Yahoo! is not affiliated with the authors of this page or responsible for its content.
DARPA Grand Challenge Technical Paper for Team CyberRider Team Leader Ivar Schoenmeyr 949 310 4255 cell







DARPA Grand Challenge Technical Paper for Team CyberRider
Team Leader Ivar Schoenmeyr
949 310 4255 cell

























Disclaimer: The views, opinions, and/or findings contained in this paper are those of the
participating team and should not be interpreted as representing the official policies, either
expressed or implied, of the Defense Advanced Research Projects Agency or the Department
Defense. DARPA and DoD cannot guarantee the accuracy or reliability of the information in
this paper. The paper is published by DARPA as a service to those who seek additional
technical information concerning the DARPA Grand Challenge of March 2004.

1

Summary: A robust vehicle with long travel suspension enables traversing rough terrain at high speeds.
The main sensing unit a dual beam sweeping laser, easily detects the edges of obstacles and terrain
gradients in real time, regardless of vehicle pitch and roll, by differentiating each beam return signal
sequentially and comparatively.

1. System Description:
Figure 1:
a. Mobility:
1. Ground contact:
Four wheels with pneumatic rubber tires. Rated
tire diameter is 39" front, 44" rear. Ground
contact area (8x10;10x12) is shown in
rectangles. (unit of measure: inches)


2. Locomotion:
A six cylinder internal combustion engine (GM V6 3.1L) provides power thru a 3 speed automatic transaxle
(P-R-N-D-2-1) The transaxle outputs to 2 drive shafts, each having 2 constant velocity (CV) joints. Each
drive shaft is connected to a drive sprocket on a bearing supported shaft, anchored to the frame, which is in
line (concentric) with the pivot of the rear trailing arm. A driven sprocket is bolted to the hub axle, turning
the rear wheel. A chain on each side transmits power between the sprockets (similar to a motorcycle).
Steering: Howe quick turn power steering rack connected to front wheels thru 2 tie rods. Front wheels are
supported by pivoting knuckles with long arm suspension members. Rear axle incorporates minimal roll
steer geometry but no active steering.
Brakes: Dual hydraulic ABS system, front and rear vented disc brakes.
3. Actuated mobility components:
Steering: Hydraulic, modified Howe power rack with custom torsion bar controlling hydraulic valve, actuated
by Hitec power servo.
Brakes: Customized Hydratech hydro boost system actuated by stepper motor lead screw, with parallel
mechanical linkage to emergency brake air cylinder to provide power off fail safe.
Shifter: Pneumatic cylinder, two 3-way solenoid valve with flow restrictor, six position limit sensing stops to
provide Park, Reverse, Neutral, 3,2,1 gear selection
Throttle: HiTec power servo; Ignition bus: electric relay; Starter motor: electric relay
Main fuel valve: electric solenoid

b. Power:
1. Power sources:
For propulsion and battery charging systems: Internal combustion engine (GM 60V6, 3.1L) equipped with
140 Amp alternator. One 12V starter battery, 2 x 12V auxiliary batteries, 1 smaller 12VDC battery for
auxiliary disable circuit. Switching power supply will provide 5V and 3VDC. A 10 pound CO2 cylinder
provides gas pressure to actuate the transmission shifter, emergency brake system and other pneumatic
components.
2. Power Consumption:
Approximate peak propulsion system power 140kW, Navigation system 1.5 kW.

2
3. Fuel type:
33 gallons of Propane, nominal.

c. Processing:
1. Hardware:
The computational systems on-board the vehicle are divided into several functional blocks, each aligned to
a particular system or task, and interconnected via Ethernet in order to send and receive commands and
sensory data. The architecture follows a Sense-Model-Plan-Act model for autonomous vehicle control,
with adaptations to provide for data access and logging across these functional blocks:

System Purpose
Hardware
PROPULSION
(single node)
Dedicated control and supervision of vehicle
locomotion, including engine, braking and steering.
Responds to commands issued by the DRIVER
node within the vehicle safety and performance
envelope. Implements the response to a hardware
E-Stop.
Implemented on a single CPU,
currently specified as an Intel
IXCDP1100 board running QNX
Neutrino. 512MB SDRAM
SENSOR
(multiple nodes,
currently
specified at
two)
Management and data processing for one or more
sensor packages (Ladar, GPS, compass,
etc.). Includes hardware interface to the
sensor(s).
Implemented using VIA Mini-ITX
small-footprint PC motherboards,
with EDEN ESP 1GHz processor
running the GNU/Linux operating
system. 512 MB SDRAM
SUPERVISOR
(single node)
Central node in a star topology for communication,
collects, logs and distributes information from
multiple sources. Also performs high-level
watchdog functions for the system.
Implemented using VIA Mini-ITX
small-footprint PC motherboards,
with EDEN ESP 2GHz processor
running the GNU/Linux operating
system. 1GB SDRAM
NAVIGATOR
(single node)
Repository for terrain database, including digital
elevation maps, vector representations of
environment. Pre-loaded with available data, this
node also collects and records sensed navigation
data. Aligned to the Plan functional block, this
node performs the macro navigation tasks.
Implemented using VIA Mini-ITX
small-footprint PC motherboards,
with EDEN ESP 1GHZ processor
running the GNU/Linux operating
system. 1 GB SDRAM
DRIVER
(single node)
Responsible for sensory fusion and high frequency
decision making, this node issues commands to
the PROPULSION node. Part of the Plan
functional block, this node is responsible for the
real-time, micro navigation tasks.
Implemented on a single CPU,
currently specified as an Intel
IXCDP1100 board running QNX
Neutrino. 512 MB SDRAM
VISION
(single node)
Gathers and processes camera data into digested
data products for the DRIVER and NAVIGATOR
nodes, primarily obstacle avoidance, lane following
and surface characterization information.
Implemented using VIA Mini-ITX
small-footprint PC motherboards,
with EDEN ESP 2GHz processor
running the GNU/Linux operating
system. 1GB SDRAM
All nodes use compact Flash cards for hard memory.

3
A separately powered independent stand-alone PC board with hardwired logic monitors the hard E-stop
binary signal output and the manual E-stop switches and implements the disable mode (communicating
directly with affected actuators) in case of a commanded hard E-stop.

2. Control methodology:
Following a Sense-Model-Plan-Act cycle, sensor data is gathered by SENSOR nodes and digested and/or
filtered into a data product for communication to the SUPERVISOR node. This data is supplied to Model
nodes (DRIVER and NAVIGATOR) for time-offset correction and integration into a pose estimate and
environment model. The environmental model is based on a discrete occupancy grid, each grid step being
.000001 nautical degrees or 4.4 inches. The grids horizontal and vertical axis represents latitude and
longitude, allowing direct correlation to waypoints tracks etc. Each grid location is assigned a value that
represents the probability that it is occupied by an obstacle. Initially all grid values are set at 50%, except
the course boundary which is locked at 100%. As the sensors scan the surrounding terrain, particularly in
the direction of the planned path, the grid values are adjusted. The objective is to clear the grid points in
the planned path as quickly as possible so it can be traversed. The dual beam scanning laser is perfect for
this task. After clearing the points immediately in front of the vehicle the laser is aimed progressively further
and further away till a distance is reached compatible with the planned speed and related avoidance
maneuvers. By continuously comparing the distance of adjacent points, comparing the distance between
contemporaneous signals from the 2 beams, and differentiating the distance between each sweep,
comparing to expected values for flat terrain, changes in the terrain due to the terrain itself or obstacles
can be detected within 25 mS. Progressive scans yield information of the (vertical) size of the threat. The
radar complements the Ladar by giving information about larger objects further away, plus moving objects
in a wider field of view. Likewise, the sonar sensors and the tactile sensors yield information of objects
close by for slow, precise speed navigation. The stereo camera helps further