Get ready, get set, race! Position and orientation data for autonomous navigation

GPS World, Sept, 2006 by William "Red" Whittaker, Louis Nastro

Preplanning information about terrain is as important as real-time navigation for achieving peak performance in autonomous driving. Both preplanning and navigation--and key technologies to support them--helped the Carnegie Mellon Red Team successfully guide the robot vehicles Sandstorm and Highlander through the 2005 DARPA Grand Challenge course.

**********

Peak performance in autonomous driving can be achieved through pre-planning information about terrain as well as real-time navigation. Both of these rely on accurate, trustworthy estimates of position and orientation. Pre-planning uses position-registered terrain data to detail an intended route and set desired speeds to achieve elapsed times that are otherwise impossible. Navigation and path tracking also rely on position estimation for effective driving.

In this article, we describe the technologies we used to guide Sandstorm and Highlander, two robots from the Carnegie Mellon Red Team, through the 2005 DARPA Grand Challenge 132-mile desert course. Our research addresses the problems of how to achieve reliable and repeatable positioning data and how to use the data to maximize the performance of autonomous vehicles.

The fusion of data from various onboard sensors provides accurate real time perception--the more data available at a decision moment, the greater the chances of a successful outcome. More specifically, effective data fusion for position and orientation estimation must persist faithfully, particularly for driving control functions, and even during prolonged occurrences of sensor outages such as blocked GPS signals. We used a Position and Orientation System for Land Vehicles (POS LV) for both preplanning and real-time operation of the vehicles. The POS LV system produced highly accurate results that were crucial to maximizing the performance of the autonomous vehicles.

Robust positioning--the ability to maintain accurate position information even during GPS outages--is a critical component for successful autonomous vehicle navigation. Accurate orientation of the vehicle is equally necessary to derive precise measures of vehicle dynamics for both pre-planning functions and real-time navigation. In either situation, it is essential for onboard sensors to know precisely where they are pointing as well as where they are located geographically in order to steer an autonomous vehicle along its intended track and negotiate past unanticipated conditions along the way.

The POS LV System

The POS LV system is a tightly coupled inertial/GPS system (FIGURE 1). Tightly coupled implementation optimally blends the inertial data with raw GPS observables from individual satellites (ranges and range rates). In this way, should the number of visible satellites drop below four, the inertial navigator will still be aided by the GPS. The result is significantly improved navigational accuracy when compared to the free-inertial operation.

Another advantage of tightly coupled integration is the improved re-acquisition time to recover full real-time kinematic (RTK) position accuracy after satellite signal loss. The inherent benefits of tightly coupled data blending become readily apparent in the accuracy and integrity of the resulting navigation solution.

By contrast, loosely coupled implementation blends the inertial navigation data with the position and velocity output from the GPS. If the number of visible satellites is sufficient for the GPS to compute its position and velocity--four or more satellites--then GPS position and velocity are blended with the inertial data. Otherwise, if the GPS data isn't available, the system will operate without any GPS aiding.

The inertial navigator computes position, velocity, and orientation of the inertial measurement unit (IMU). The Kalman filter estimates the errors in the inertial navigator along with the IMU, distance measurement instrument (DMI), and GPS receivers. The only addition to this system setup was a Trimble Ag 252 receiver, which provided Radio Technical Commission for Maritime Services (RTCM) standard corrections from a virtual base-station (VBS). Typical position accuracies for open-sky conditions are in the order of 0.5 meter root mean square (RMS).

The GPS Azimuth Measurement Subsystem (GAMS) integrates the IMU with a two-antenna heading measurement system. As long as there is GPS coverage, GAMS continuously calibrates the IMU and the azimuth does not drift. A single-antenna configuration, in comparison, requires dynamic heading alignment and delivers heading measurements that suffer from drift.

GAMS uses a carrier-phase differential GPS algorithm to measure the relative position vector between the two antennas. GAMS uses carrier-phase measurements from five or more satellites to estimate and, eventually, to identify a set of integer phase ambiguities for each satellite being tracked by both receivers. For the ambiguity resolution algorithm to work, both receivers must track at least five common satellites. Once tracking has been obtained, GAMS will continue to operate with as few as four satellites. The GAMS heading system will not provide measurements when fewer than 4 GPS satellites are available. During GPS outages, POS LV will continue to provide accurate heading measurements drifting at the rate of about 1 arc min/min. Accurate heading is critical for robotic vehicle navigation, especially in intermittent GPS conditions.