, Assistant Professor, Computer Science and Engineering
The metro region contains a large number of traffic lights. When one or more components in a traffic light system fail, the system will malfunction. A malfunctioning traffic light is at the very least inconvenient, and at the worst dangerous. In the near future, a large fraction of the traffic lights will possess advanced diagnostic capabilities. This capability may be summarized as follows. The operator of the traffic light has the capability to define normal operating conditions. The traffic light controller will possess the ability to report remotely to a maintenance engineer when the light is operating abnormally. Apart from faults, abnormal traffic patterns can also cause abnormal operation of a traffic light. An engineer has to sift through several pages of 'event reports' and identify those reports which actually correspond to faults. Processing the event logo is highly labor intensive and time consuming. The goal of our research is to develop new techniques to process event logs recorded each day to distinguish actual faults from false alarms. The analysis will account for the history of the system in both time and space. Detection techniques used to detect faults in multiprocessor systems will be our starting point. If the first phase of the research is successful, we will attempt to develop diagnostic techniques to potentially identify the cause of failure as well. Our objective is to develop a software system which will assist an engineer in processing the daily event logs and identifying faults. The successful completion of our research will reduce the time spent in and lower the difficulty of maintaining the traffic control system.
- Project number: 1995011
- Start date: 06/1995
- Project status: Completed
- Research area: Transportation Safety and Traffic Flow
Maintenance, Traffic operations