Periodically assess the accuracy and reliability of parking management vehicle sensors and predictive algorithms to assure performance under a variety of weather conditions and traffic situations.

Experiences from deploying a pilot parking management system in Cape Cod, MA

Date Posted
08/30/2017
TwitterLinkedInFacebook
Identifier
2017-L00765

Cape Cod National Seashore Parking Management System Pilot Synthesis

Summary Information

The Cape Cod National Seashore (CACO) has undertaken a program to improve parking management at its beach parking lots and to provide information about parking availability to CACO visitors. In summer of 2013, as a first step toward deployment of the full Parking Management System (PMS) , the Volpe Center installed a pilot system at the Little Creek parking area that serves Coast Guard Beach. The pilot system was in operation for the duration of the summer season (through September 30).

Key components of the parking management system included:
  • Two passive, magnetometer-based vehicle sensors (inbound and outbound) to record passing vehicles as they enter and leave the parking area.
  • A software module that uses inbound and outbound vehicle count data from the vehicle sensors to track parking availability, and to predict what time the lot will be full. When deployed, the full parking management system will use this predictive information to issue "lot full" notices in advance of a given parking area actually being at capacity
  • A video camera and digital video recorder (DVR) to corroborate the data the vehicle sensors collected. The video camera captures time-stamped images of the parking lot at a rate of 1 frame per second, which are periodically downloaded onsite from the DVR and compared with sensor data from the same dates.

Lessons Learned

Key Lessons from the operational period were as follows:

  • Verify that magnetometers can accurately record vehicles entering in rapid succession to prevent "under-counting"

    Performance of the vehicle sensors is the key to the system’s performance. It was discovered early in the field test that the magnetometers were not functioning as well as expected due to the nature of the traffic. Observation revealed the magnetometers did not accurately record vehicles that were closely spaced and entered the parking area in rapid succession (tailgating). This occurred in cases where several successive vehicles had season passes and therefore did not need to stop at the entrance booth (the Rangers waved them through). The vehicle sensors require several seconds between vehicles in order to reset themselves, so when there were several vehicles entering in rapid succession, the entrance sensor recorded the first vehicle, and typically missed the subsequent vehicles. This &successive vehicle” problem is potentially significant; strings of successive vehicles with season passes are common during CACO’s peak summer season. The problem also could occur at the outbound sensor, depending on the volume of exiting traffic
  • Locate the sensor several feet past the entrance booth to prevent "double-counting"

    The performance of the magnetometers was affected by their placement. At the pilot site, traffic cones were used to narrow the entrance and exit lanes slightly to ensure that vehicles pass close enough to the sensors. If magnetometers are used in the final installations, it may be desirable to embed the sensors in the pavement so that vehicles pass directly over them.



    In a limited number of cases the pilot system’s inbound sensor double-counted vehicles. This occurred in cases when a vehicle “overshot” the entrance booth, and pulled to a stop past the doorway. In these cases, the vehicle triggered the sensor. Because magnetometers are motion sensitive, once the vehicle stopped moving the sensor reset itself. Then, when the vehicle pulled away from the booth it triggered the sensor again, and the system counted the vehicle twice. This issue would be resolved by locating the sensor several feet past the entrance booth. Vehicles would trigger the sensors only after they had completely cleared the booth.
  • Pick a Sensor type that is best suited for your use

    Since all sensors have unique strengths and weaknesses, the Volpe Center decided to replace the magnetometers during the field test. On July 29, 2013, the Volpe Center installed two sets of breakbeam sensors. The sensor chosen was the Dakota BBA-2500 wireless breakbeam system. Its features include a design that is intended for outdoor use, an integrated solar charging system and a transmitter that enables the sensors to be located up to a half mile from the receiver with no cabling. Using breakbeam sensors by themselves at the fee booth would not work well because rangers, beachgoers and others sometimes walk or stand outside the booth and could repeatedly trip the sensors. Because the roadway leading to the Little Creek parking facility is divided in some places, it was possible to place the sensors about 150 feet from the fee booth and still achieve lane separation. The sensors use a single 4-channel receiver which replaced the two magnetometer controllers inside the fee booth. The normally open relay contacts were connected directly to the DVIF-10 which sends RS-232 signals to the computer. Performance of these sensors seemed to more accurately match the fill rates of the lots as seen by the video.
  • Incorporate a Video Camera to provide a measure of confidence in the system

    The video camera that was installed by Volpe to provide a measure of ground truth to corroborate the performance of the system, proved to be a valuable tool. Access to live images from each of the parking areas should be incorporated into the final system if possible. Live video can achieve ancillary benefits as well, including staff safety and forensic information if recorded.
  • Take into consideration the known fill patterns of a particular lot during algorithm development

    The accuracy, and in fact the need for the fill rate prediction algorithm came into question over the course of the field test. Traffic tended to arrive in waves, which caused the anticipated fill time of the lot to change significantly over the course of the day. In looking towards the next phase, there are many other ways in which this algorithm might be approached by taking into consideration the known fill patterns of a particular lot. Experienced rangers could tell you that if a lot hasn’t filled by a particular time, it never will. Temperature, cloud cover, day of the week, shark threat and point in the season all play a role in the demand for parking, and there are ways these might be taken into account. Since beachgoers generally only care if a lot is currently OPEN or FULL, it may be sufficient for the PMS to know simply that a lot has reached 95 percent of its capacity, at which time travelers might be directed to other facilities. As this project enters its next phase, other approaches to handling lot fill rates will be considered.



    In looking toward the next phase, it makes sense to work towards a system whose vehicle counting technology is adaptable to a variety of situations. In addition, there were reliability problems associated with using a Windows based computer in this environment. It makes sense to build the final system using a programmable logic controller that processes the data locally, and is more tolerant of high temperatures, humidity and autonomous operation. There are intelligent video-based systems that can be customized to specific scenarios, and can be programmed to process the data internally. The Volpe Center will investigate the potential for using these systems because they may resolve all of the issues encountered during this field test.