Design and tailor system technology to deliver information of useful quality and quantity, that the user can reasonably absorb.

Virginia's experience implementing a pilot automatic vehicle location (AVL) system in an urban winter maintenance operations setting

Date Posted

Lessons Learned From a Pilot Project of an Automatic Vehicle Location System in an Urban Winter Maintenance Operations Setting

Summary Information

The Virginia Department of Transportation (VDOT) maintains all interstate, primary, and secondary roads in 93 of Virginia's 95 counties. Snow removal and ice control activities are of special concern in urban counties since they slow traffic on an already congested highway system and adversely affect a large part of the population.

VDOT instituted a pilot automatic vehicle location (AVL) system to attempt to improve the management of operations and communications during emergencies. AVL is a technology that locates vehicles using a global positioning system (GPS), monitors vehicle activity, transmits vehicle location and activity information to a remote site, and displays the information on geo-referenced maps. As part of the pilot project, VDOT installed GPS units in 80 trucks. The pilot project extended over three winters, from 1997 through 2000.

The purpose of the project was to evaluate the AVL system's effect on (1) the administration of snow removal and ice control contract forces, (2) the provision of information concerning road conditions, and (3) the management of snow removal and ice control activities. The system tracked vehicles to an acceptable degree of accuracy. However, due to operational and institutional issues, system problems, and mild winters during the study, no financially quantifiable savings could be determined.

VDOT's experience in implementing a pilot program for automatic vehicle location (AVL) system in its snow removal truck fleet provides some useful guidance for this type of deployment. The system should not be cumbersome to use so that the truck drivers were not overwhelmed. It should be tailored to deliver information of useful quality and quantity that the drivers can easily understand and absorb.

  • Use detailed and appropriate background mapping: VDOT chose two geographically accurate and easily readable base maps for the AVL system; one was based on aerial photos and the other on street vectors. The aerial photos are particularly useful in presenting cultural data references not visible on street vector maps. For example, a vehicle may appear to be far off the road when displayed on a street vector map but may be seen actually to be in a parking lot turning around when displayed on an aerial photo map.
  • A map dataset for the AVL application was compatible with other GIS applications. This allowed distribution of the cost and permitted more options for shared use and common display formats for geospatial data. The maps used in AVL should be useful across organizational boundaries, e.g., monitoring by the Safety Service Patrol, police surveillance, highway maintenance, land use, and tax parcels.
  • Upgrade the system as necessary for better data collection: The AVL system tested did not perform to expectations until the third winter season. Frequently, the accuracy of the AVL system did not meet the 5-meter (15 foot) requirement. The average accuracy was approximately 10 to 15 meters. By the third year, this was acknowledged to be sufficient to locate and track the vehicles. The desire to discern which specific lanes were traversed was acknowledged as more information than could be absorbed during real-time operations.
  • A design flaw in the tracking feature caused the system to slow down as the volume of stored vehicle tracks increased. During the first two winter seasons, the location and sensor data for each in-vehicle unit (IVU) were collected every two seconds and transmitted to the base computer every ten seconds. The first and second winter, the speed of the system update became unacceptable after approximately one hour of data collection. This problem was resolved during the third winter by a combination of system improvements and decreasing the data collection rate to once every ten seconds per IVU (and transmitted to the base computer once every minute). During the first two years, the failure of the system to meet the expectations of supplying real-time information on location and activity on a regular basis undermined the confidence of the field personnel and resulted in little reliance on the system at the area headquarters level.
  • Design the system with only the necessary details and do not complicate the design: The contract specification called for the AVL system to generate a real-time display of vehicle location, trace the route of each vehicle tracked, and color-code the traces to identify the activity performed. Although the information display of the AVL system met these requirements, the managers and monitors did not use the color-coded information for real-time management decisions. They were interested in the vehicle location but did not rely on the color-coded track, which represented activity. This appeared to be more information than they could absorb and react to, considering the number of vehicles being monitored. The AVL system archived the activity monitored by the sensors. Although some inaccurate information was recorded, this was determined to be due to faulty wiring of the sensors, not the AVL system. No procedure to tie performance, or compensation, to this activity data was considered.
  • Define specific use of different systems, i.e. two-way messaging: Two-way messaging was included in the AVL system as a means of directing the contractor portion of the snow fleet that was not on the VDOT radio network. The intention was for supervisors at headquarters to send command messages to driver IVUs (start plowing, return to base, etc.) and that drivers would be able to send a limited set of canned messages back to headquarters (finished plowing, I'm lost, yes, no, etc.). In practice, however, two-way messaging was of lesser value to the contractor fleet than to the VDOT state fleet, even though VDOT trucks also had radio. Although the same hired trucks were used during each storm event, the operators employed were often different. The turnover in operators negated the operator training undertaken prior to the winter season and burdened the staff responsible for the AVL system with training the new hired truck operators in the use of two-way messaging as they reported for duty at the start of the storm. This proved unworkable and led to the abandonment of the AVL communication system as the preferred form of communication. For contract drivers, AVL messaging proved too complex and mostly unusable.
  • On the other hand, there were numerous anecdotal examples of drivers who knew the system of using two-way messaging with great efficacy. A simple rule is that if one has regular daily users who are quite familiar with the system, two-way messaging can be a significant benefit. One lesson learned was that two-way messaging must be tailored to the user. Although two-way messaging is an important tool, requiring operator responses or inputs while a vehicle is moving is unreliable. Further, the report recommended that two-way messaging be used only in the case of an emergency since it diverts equipment operators from their primary objectives.

This lesson documents the importance of designing useful data outputs in implementing new technologies. As seen in these examples, it is important to keep the design simple, by only including necessary components; it's not efficient to have a system with high-end technologies if the audience will not use them, and in the end the users' needs and abilities should always be kept in mind when deploying new systems. On a more detailed level, it was noted that 1) good background mapping is extremely important, 2) the determination of lane location is impractical during real-time operation, 3) collection of data on each vehicle every two seconds is not necessary since lane location information is not needed, 4) there is a limit to the amount of sensor information that the system monitors can absorb in a real-time situation.