DOO: what method do you prefer?

[HSTEd]

With DOO, once the unit moves then it's in the hands of the Gods. At least with a Guard or Platform staff you have a chance.

It's probably possible to use object detection (from cameras, LIDAR, RADAR et al) technology to determine there is a person in contact with the train who is mysteriously moving at exactly the same speed as the train as it pulls away.

Not many trains have people sized objects stuck to the side of them after all.

 

[TurboMan]

It's superfluous. The train safety check takes place BEFORE the train moves. Once the train is moving then all bets are off.

With DOO, once the unit moves then it's in the hands of the Gods. At least with a Guard or Platform staff you have a chance.

That depends on whether there are in-cab monitors and at what point they deactivate.

 

[Bletchleyite]

It's superfluous. The train safety check takes place BEFORE the train moves. Once the train is moving then all bets are off.

It's like that now, it doesn't have to be. I get the driver distraction bit, but AI tech offers an interesting option.

 

[ComUtoR]

It's like that now, it doesn't have to be. I get the driver distraction bit, but AI tech offers an interesting option.

It isn't just about distraction. You are supposed to prevent incident, not track it in full HD. Once you take power, its too late. That person gets dragged down the platform and often has "life changing" injuries.

I have seen some very interesting camera tech that detects passengers and objects at doors but again, the theory is to prevent an incident from occurring. Trap and drag shouldn't happen and we should be doing everything to prevent it. Everything that happens after the incident occurs is just mitigation.

AI tech is very interesting but just like driverless trains; its *light years away.


== Doublepost prevention - post automatically merged: 14 Dec 2022 ==

That depends on whether there are in-cab monitors and at what point they deactivate.

I noted you posted the RSSB link but have you also read any court cases regarding PTI incidents ?











*yes, I know

 

[Bletchleyite]

It isn't just about distraction. You are supposed to prevent incident, not track it in full HD. Once you take power, its too late. That person gets dragged down the platform and often has "life changing" injuries.

But that's like "why make trains crashworthy? We just stop them crashing."

If prevention fails, then comes mitigation.

 

[ComUtoR]

But that's like "why make trains crashworthy? We just stop them crashing."

If prevention fails, then comes mitigation.

I would agree.

With DOO and dispatch, the better options are prevention. We have PTI policies, Rules, strategies, and some interesting technology all based on prevention.

PTI mitigation is interesting for sure and should absolutely be explored. As a Driver, I want to do everything I can to prevent an incident from taking place. Being told you dragged the passenger a few yards down a platform and they only lost their legs isn't a pleasant experience. As a DOO Driver my thoughts always lean to why it happened in the first place. It literally only takes a few seconds and a centimetres of movement and a passenger can lose a limb. I would rather spend those AI millions on prevention.

 

[zwk500]

It isn't just about distraction. You are supposed to prevent incident, not track it in full HD. Once you take power, its too late. That person gets dragged down the platform and often has "life changing" injuries.

I have seen some very interesting camera tech that detects passengers and objects at doors but again, the theory is to prevent an incident from occurring. Trap and drag shouldn't happen and we should be doing everything to prevent it. Everything that happens after the incident occurs is just mitigation.

AI tech is very interesting but just like driverless trains; its *light years away.

I don't think it's as far away as you suggest - The recognition algorithms already exist, and in some DOO setups the monitors remain active after closing the doors. It's not hard with modern TMS systems to link the algorithm to the power controller and if the system detects a trap it sounds an alert and prevents power being taken, forcing a double check.

Whether this is a reasonable reaction considering the likelihood of an incident occurring, the severity when it does, and the impact on everybody else of delays from both false and actual positive alarms is a different discussion.

 

[ComUtoR]

I don't think it's as far away as you suggest - The recognition algorithms already exist

I've seen some pretty awesome stuff.

and in some DOO setups the monitors remain active after closing the doors.

Yep, I drive those.

It's not hard with modern TMS systems to link the algorithm to the power controller and if the system detects a trap it sounds an alert and prevents power being taken, forcing a double check.

One of my biggest issues with the railway is that tech is poorly implemented or takes second place to costs. What also drives me to distraction is when someone says 'its not hard' or 'tech can solve it..' I have spoken directly to people who are tying to put this kind of technology on trains and are directly involved with on board tech. It just isn't happening right now. There are also lots of technical briefs and research papers available online. This is an area where I want to be wrong and I want all this cool tech in place as fast as possible.

Whether this is a reasonable reaction considering the likelihood of an incident occurring, the severity when it does, and the impact on everybody else of delays from both false and actual positive alarms is a different discussion.

PTI risk is pretty much top of all railway risks. At my TOC there has been quite a few PTI incidents. Not all severe but enough to make you seriously consider the true risk of DOO, and indeed all methods of dispatch.

 

[zwk500]

One of my biggest issues with the railway is that tech is poorly implemented or takes second place to costs. What also drives me to distraction is when someone says 'its not hard' or 'tech can solve it..' I have spoken directly to people who are tying to put this kind of technology on trains and are directly involved with on board tech. It just isn't happening right now. There are also lots of technical briefs and research papers available online. This is an area where I want to be wrong and I want all this cool tech in place as fast as possible.

Good to know, and I share your frustration at the slow pace of progress, or when costs/bad specs are allowed to get in the way of otherwise good ideas. So much work that absolutely could be being done by computers (we know because other railways are already doing it) is held up by organisational lethargy or squabbling over which budget it gets allocated to.

 

[class68fan]

More interesting working is at Banbury station if from platform one and two going to Bicester on a chiltern railways turbo 165or turbostar 168 service train is dispatched. Or a gwr 165 to Didcot Parkway also dispatched with a dispatcher. With dispatcher tipping the Driver and Driver looking out. If it is a 800 802 of platform two dispatcher shows light to driver for dispatch via the trains CCTV cameras for doo dispatch. Note no dispatcher the evening class 800 802 service from platform two to London Paddington starts from Oxford.
Yet platform four at Banbury for local Chiltern services and Gwr services to Didcot Parkway and London Marylebone respectively is the Doo CCTV screen.
Chiltern 165 and Chiltern 168 to London Marylebone can use platform two and so do Gwr 165 turbos and quite often a look back occurs. Note on the chiltern railways services after the Train Manager gets off they do assist the driver by remaining at the rear whilst driver looks out. Banbury is quite a interesting station for the above dispatch matters. Also from the platforms at Banbury sometimes the gwr turbo drivers use the cctv from the turbo as well. Interesting workings at Banbury.

lot wrong in that post

YOu see them doing it at the MET stations and they are not supposed to

on look back on met lot wrong in that post