On February 2006, a contract for testing TrafficSense was signed between Kansas DOT and Cellint to compare road traffic data calculated from cellular network data by the TrafficSense system to the existing road sensors data of the SCOUT system in KC (www.kcscout.com). The system was deployed within eight weeks, and after few more weeks of calibration became operational.
Changes in speed of the selected roadways were measured on data for 44 days (6/18/06 to 7/31/06) and compared to the road sensors, with special emphasis given to slowdowns of traffic.
Slowdowns are defined as a decrease of speed by at least 10 MPH during less than 10 minutes, where the speed before the slowdown is higher than 50 MPH and the speed during the following 10 minutes drops below 50 MPH.
Summary:
The table in the next page summarizes the comparison analysis between TrafficSense and the inductive loops at the SCOUT traffic control system at Kansas City. The analysis was performed on “blind test” data: Cellint supplied KDOT/KCScout with data during June and July 2006 and received the comparable sensor data only in August 2006. 36 files of raw data are attached to this report, together with comparison graphs such as the samples below (Magenta for road sensors, Blue for TrafficSense). The analysis shows that the average latency of detecting slowdowns by TrafficSense in comparison to road sensors at the location of the sensors is about 4 minutes, and the average difference between the systems in measuring the local speed over the sensors is less than 5 miles per hour.
Taking into account few other factors, as detailed in the comments below, will show that the real latency at the sensors’ location is shorter.
Graph Samples:
Findings:
The analysis shows that:
There is a very high correlation between the cellular data generated by Traffic Sense system and those of the existing loop road detectors:
- The average latency of detecting slowdowns by TrafficSense in comparison to road sensors at sensors’ location is about 4 minutes*
- The average difference between the systems in measuring the local speed over the sensors is less than 5 miles per hour
- The overall number of false alarms of the sensors was significantly higher than TrafficSense’s
- System uptime was more than 99%
* KDoT/SCOUT may use different criteria for slowdowns, so their final results may be different.
Observations:
- TrafficSense proved to be a reliable and accurate traffic data collection system in real time.
- This is the first cellular based traffic data collection system in the US to be successful at a controlled pilot, with results that were verified by an independent examiner.
- Deployment and calibration of TrafficSense are much faster than traditional road sensors and don’t require dangerous road work at construction zones which contribute to traffic delays.
- TrafficSense’s cost benefit ratio suggests a much better return on investment for traffic agencies and road operators.