The I-880 Field Experiment:
Effectiveness of Incident Detection Using Cellular Phones
Alexander Skabardonis, Ted Chira-Chavala, Daniel Rydzewski
ABSTRACT
This report describes the evaluation of the effectiveness and adequacy of cellular phones for incident detection as an alternative to infrastructure-based surveillance systems. The analysis was conducted as part of the I-880 field experiment using the California Highway Patrol's (CHP) Computer Aided Dispatch (CAD) incident database. Cellular phones have the highest detectionrate among the detection sources examined. They detect 38 percent of the freeway incidents (accidents and lane-blocking disablements). The combined cellular phones, freeway service patrol (FSP) and the CHP detect 75 percent of all the incidents. The results from the statistical analysis indicate a significant effect of the incident detection source on the incident duration. Incidents reported by cellular phones show greater incident durations by an average of 14 minutes than similar incidents reported by the CHP or the FSP. This additional delay is due to the incident verification process.
EXECUTIVE SUMMARY
Objectives and Methodology
Unpredictable events that occur frequently on freeways include accidents, stalled vehicles, and spilled loads. When incidents occur during peak periods and affect the available freeway capacity, motorist delays usually increase many-fold. Freeway surveillance and incident management systems are key components of the Advanced Traffic Management and Information Systems (ATMIS) aim to detect and respond to incidents reliably and rapidly. One disadvantage of infrastructure-based freeway surveillance systems is that they are capital intensive due to the high costs of installing inductive loop detectors, closed-circuit television (CCTV) and trunkline communication along the entire roadway.
Cellular phones have served as sources of incident detection in the San Francisco Bay Area. Calls can be made from cellular phones by motorists to report incidents on freeways. Information obtained from cellular phones vary in the detail and quality, and also, the incident may be reported after considerable time has elapsed. Therefore, the feasibility of freeway surveillance systems utilizing cellular phones needs to be carefully evaluated. This report presents the findings of the evaluation of the feasibility of using cellular phones for freeway surveillance as an alternative to infrastructure-based surveillance systems.
Incidents reported by cellular phones and other sources on a 9 mile section of the I-880 freeway were obtained from the California Highway Patrol (CHP) computer aided dispatch (CAD) center, along with field observations on incidents by probe vehicle drivers traversing the same freeway section with an average headway of 7 minutes. Supplementary data were gathered from the freeway service patrols (FSP) and tow truck companies' logs. The database includes 264 incidents (accidents and lane-blocking vehicle disablements) and 1429 "other events" (vehicle stalls, and other non-accident events that did not block travel lanes.)
The data were analyzed to assess the quality and adequacy of cellular phone calls for incident management using the following performance measures: a) incident detection rate, b) false alarm rate, c) timeliness of incident detection, and d) available details essential for initiating response actions for incident removal (incident location and type, incident severity, and the number and type of vehicles involved). Analyses also were performed to assess the effects of the timeliness of incident detection on duration, and the effects of incident duration on traffic congestion.
Findings
Cellular phones detect 38 percent of the incidents, and 1 percent of the other events. This is probably because incidents (having blocked travel lanes and impeded traffic flow) are likely to get immediate attention from other road users. On the other hand, other events (breakdowns on the shoulders which occur with greater frequency and do not present a hazard to other motorists) are often not noticed by other motorists. About 7 percent of all reported incidents by cellular phones are false alarms (moving violations and other events that could not be verified by the CHP). The false alarm rate for other events is much higher (32 percent), reflecting cellular-phone callers' difficulties in judging whether vehicles restingon freeway shoulders are disablements or just temporary stoppage. A significant contribution of cellular phones is that they capture additional 35 percent of the incidents not witnessed by the CHP. The detection rate by the combined CHP and cellular phones was 60 percent. Adding FSP to the combined CHP and cellular phones increases the incident detection rate from 60 to 75 percent, i.e., the majority of the freeway incidents were detected by the CHP/FSP (that are typical components of urban freeway management) plus the cellular phone users.
The evaluation results show that relative to other detection sources, cellular phones have the highest detection rate and are the fastest detection source. They are reasonably effective (with above-average ranking) in terms of correct reporting of incident locations, and availability of information about the incident type and the number of vehicles involved. Weaknesses of cellular phones include a very low rate of detecting other events, the highestrate of false alarms, and limited information on the incident severity. Also, cellular phonesneed verification and cannot tell when the incident is cleared.
The testing of loop data based incident detection algorithms produced very low incident detection rates and high false alarm rates. Reasonable performance of these algorithms was reported when they applied to only a small preselected sample of lane-blocking incidents. These results indicate that existing incident detection algorithms do not perform satisfactorily in real-world operating conditions with high frequency of incidents and other events.
Incidents reported by cellular phones show greater incident durations by 14 minutes on the average than similar incidents reported by the CHP/FSP. This extra delay is due to the incident verification process. When the CHP (or FSP) detects an incident, they usually take actions to respond to it immediately. On the other hand, for the incidents reported by cellular phone, the CHP officer is dispatched to the scene to verify the existence of that incident before taking response actions. The verification plus response time is significantly affected by the incident type and detection source. Smaller response time can be expected for accidents than for lane-blocking vehicle disablements, and for incidents reported by cellular phones than call boxes. The clearance time is expected to be the highest for injuries, followed by for non-injury accidents and breakdowns in travel lanes. Further, for each incident type, events that take longer to verify/respond to also take longer to clear.
Recommendations
Incident management requirements for ATMIS systems cannot rely solely on cellular phones. Cellular phone reports may contribute significantly to the incident detection in combination with other sources, and may be utilized in the verification of incidents detected by loop based systems. This would require proper fusion of cellular phone data with information from other sources and utilization of other technologies (video surveillance.)
Under the existing incident management practice, incident detection by cellular phones (as well as by call boxes or public entities) is simply not as time-effective as incident detection by the CHP or the FSP. To make cellular-phone incident detection as effective as the detection by CHP/FSP, in terms of the incident duration, current incident verification and response practices associated with cellular-phone detection need to be revised.
Full Report: UCB-ITS-PRR-98-01 (885 KB PDF file)
Home | General Info | Current Research | Researchers and Staff | Publications and Video | What's New | Media Info | PATH Partners | Search | WWW Links | Contact PATH