1594651 - Features and Limitations of Video-based Guideway Intrusion Detection and Impact on Operation

In metro systems, up to thousands of passengers per hour are on station platforms and wait for their train. At the same time, trains enter stations with up to 40 mph. Unless there are specific protective measures, people can easily enter the station tracks or tunnels in underground systems. Due to the relatively high speed and long braking distance of a train, stopping in time is often not possible. In the extreme case, an accident with severe injuries or even a single fatality may result from such a behavior. This makes the platform and platform tracks critical zones for metro operation.

While UTO/GOA4 systems are typically equipped with Platform Screen Doors, systems of lower grade of automation, i.e. GOA2 or GOA3, do not feature such protection systems. We have therefore analyzed the potential benefits regarding safety if additional measures would be taken to detect people on tracks and warn train drivers or stop trains automatically, e.g. through CBTC-based vehicle commands.

If people enter the tracks, it is often assumed that this happens accidentally, e.g., because of a crowded platform, distraction by smartphone, or influence of drugs. However, we could observe a substantial number of people who enter tracks intentionally and leave before arrival of the next train. Nevertheless, some operators consider a sufficiently safe track supervision system necessary to cope with accidents that result from track intrusions.

This paper presents the empiric experience from a 3-year observation campaign of platform tracks in a conventional subway system. The intrusion detection system which we used as benchmark is camera-based, so it is possible to confirm visually why people entered the track. The observed “incidents” were classified according to the type/reason of the intrusion and combined with additional characteristics, e.g., day of week, time, location along the platform edge, whether intruders rather jumped or slowly slid on the track. With these characteristics, it is estimated that more than 95% of the observed intruders intentionally entered the track. By analyzing the time of train departure, time of intrusion, and arrival of the next train, some unexpected results have been obtained. Our statistics indicate strongly that most intruders are well aware of the risk they take. However, near misses had been observed, too with people just leaving the track shortly before train arrival. Taking into account the total number of incidents, including false alarms in comparison to near misses and resolved intrusions which did not result in any safety problem, the typical requirements for track intrusion detection system are worth to be re-considered. The paper concludes with some recommendations how to implement and operate track intrusion detection systems for GOA2/GOA3 subway systems without impacting availability too much but still provide a benefit in terms of safety.