This year's challenges
In anticipation of the 2005 event, Team DAD has moved to an active laser-based system instead of a stereo-vision system to guide the truck’s self navigation. This system is a 64-element rotating LADAR (laser radar) terrain-mapping and obstacle-detection system employed as the single environment sensing system. The system has a 360-degree horizontal field of view and a 20-degree vertical viewing angle, providing robust real-time 3D terrain mapping and overcoming limitations of traditional vision-based systems including rain, darkness, and fog.
Laser systems are based on light emitters and sensor diodes. To determine the distance from objects that may be in the path of the truck, the system fires a laser towards that object and monitors the time it takes for the light to return to the sensor diode, indicating distance from the truck. Repeating this process at high speed provides all the information needed to detail an extremely complete terrain map that is more accurate than any previously designed system.
This system camera employs advanced DSP techniques for multiple reflection discrimination, variable laser intensities, and redundant sub-camera assemblies, and is mounted on top center of the cab, giving it a clear view in all directions. Rotating at 600 RPM, the camera is shock mounted and has an inertial navigation system (INS) to report exact pitch and roll of the unit. These data are used by navigational computers for correction factors.
The LADAR unit generates its own light and uses a proprietary filter to reject sunlight, so it works well under all lighting conditions. Because the whole camera spins, dust and rain are spun off the unit as it rotates. The LADAR is capable of seeing through fog and heavy rain by ignoring early reflections and also incorporates a "dynamic power" feature that allows it to increase the intensity of the laser emitters if a clear terrain reflection is not obtained.
Texas Instruments C6000 DSPs process data from the LADAR, the GPS units, and INS. A DSP, running at 1GHz, analyzes the desired vehicle path and speed and reconciles the possibility of navigating that path with the LADAR terrain map. If an acceptable route and speed is not found, the system looks off-centerline for an acceptable path. The LADAR system identifies the size and distance of objects in view, including the vertical position and contour of the road surface itself. All objects are presumed solid and if they are not road surface, they are to be avoided.
View a full size image
In addition, all paths are assessed incorporating the G forces that would affect the vehicle if that particular path were followed. This information is used to determine the maximum speed that the vehicle should be traveling. Acceleration or braking commands are issued accordingly. In all cases the software seeks the best available road surfaceand thus, the fastest possible speedwithin the boundaries of the GPS waypoint being traversed.
The information provided by the laser imaging system is reconciled with expected vehicle path data. The resultant navigational control data is then sent to the vehicle servo-control navigation system, enabled by DSP-based digital signal controllers. The vehicle servo-control system actually controls the motion of the vehicle. Servo control takes its lead from the laser-based imaging system, but is also augmented by two different high precision GPS receivers, servo motors to actually navigate the truck, and an IMU (Internal Measurement Unit).
The IMU uses gyros and accelerometers to correct for gaps in GPS signals. The gyro and odometer system are adequate to navigate properly until GPS signals are reacquired while passing through a tunnel. However, the system defaults to the LADAR system anytime the GPS signal is not stable.
The computer system uses only a few watts of power and has no fans. The entire DAD navigational system is purpose-built from the ground up. Thus, all computing power is embedded and occupies a dramatically smaller footprint than conventional computing systems.
The LADAR terrain mapping and obstacle detection system was fully conceived and modeled using SolidWorks CAD/CAM system. Mentor Graphics PADS was used for the entire PC board layout. TI’s Code Composer Studio program was used for all software development. All software code was written in-house in assembler language.
DAD engineers applied lessons learned from the first Grand Challenge to improve their vehicle in the areas of tires, wiring, mounting hardware, and field-testing. Team DAD does not expect to let a rock get in its way this year: The team has coded contingency plans for deviations from the desired speed. This means that if the truck lodges against a rock, it will apply more throttle.
As the only vehicle competing in the 2005 DARPA Grand Challenge solely relying on DSP technology, the Team DAD truck’s autonomous decision-making capability can assess the road surface and potential obstacles out to over 500 feet, and do so over 60 times per second. This is enough processing power to navigate a car traveling at over 100 mpha significant competitive advantage over other sensor- and vision-based entries. Additionally, because DSPs have been in use for decades controlling mission-critical systems for commercial and government applications, their reliability is excellent.
As in 2004, DAD’s vehicle is driven as a regular car when not in autonomous mode. Autonomous mode is engaged with the three switches mounted on the roof of the vehicle just above the rear view mirror. When these switches are disengaged, the vehicle drives just like a regular truck. This "light-footprint" approach afforded by embedded DSP computers allows for a full compliment of passengers and cargo whether or not the vehicle is operating in autonomous mode.
Most individuals have fantasized at least once about a car that can drive them to work while they ignore the road, relax, and just enjoy the ride. While it may be a while before everyday consumer cars will shuttle around town while "drivers" read the morning paper, the future of autonomous vehicle technology is well on the way to being a reality sooner than most people think.
Bruce Hall is president of Digital Auto Drive