Hi James,


Apologies for my slow reply - I got distracted into other less useful discussions on this forum! (Literally don't go there!)


The problem is the very high levels of safety and reliability you need. Even if you say this is an add-on that supplements rather than replaces the driver, so that the driver is the main control of the brakes and this has additional control, there are two huge problems. One is that it must not apply the brakes when it shouldn't - emergency braking a train is considered a hazard (there have been a number of serious passenger injuries caused by it). So it must never falsly think that something is on the track when it isn't. The other is more subtle, and was found in early adoptions of Automatic Train Protection. Once the driver knows there is an automatic protection system in place they will tend to rely on it, and if it isn't fail-safe horrible things can happen... These may be solveable, but are they solveable at a cost that is less than ETCS?


What may be interesting is a simple alarm system, so if a visual system picks up what it thinks is a hazard it alerts the driver who then decides what to do about it. That would have far fewer safety implications. But given the cost of railway approvals, due to the fact that no possible risk must come from any change, it would be 'interesting' getting it supported by major authorities. (Ref my comments above about this not influencing the accidents we're actually seeing.)  If it were me, I'd start by looking at it as a tramway solution - which I'll bet someone already is!


You're probably aware that SPADs are a pure financial issue, the UK ATP system is in place for this, just needs to be rolled out more. (Or ETCS.) SPADs aren't quite as scary as the "red-tops" liked to make out in the early 2000s, there are huge protection zones in place. But they're not good of course - every time you degrade a safety margin then by definition there's a bit less left!


But regarding driverless trains generally, no problem, we've got them. And they meet the levels of safety demanded. If only we could afford them...


Rule one of railway engineering: Nothing is cheap, nothing is simple, unless you can persuade the public to compromise on safety,

Rule two of railway (and indeed safety) engineering: Every time you eliminate one Hazard, you create ten more!



It is very, very frustrating, being someone who is by instinct an innovation manager, to have to write posts like this. And as you can imagine, since I moved to the rail industry I've had many, many discussions like these. We desperately need innovations, particularly low cost innovations. But my (long!) experience has been that the only approach I've seen work is to start by working out how you're going to do it (whatever "it" is) safely and reliably at a "broad brush" level, and then work out the detailed technical solution - which is much the easier bit. Trying to start with a piece of whizzy technology and then trying to make it safe is a nightmare, indeed almost impossible. It's a strange business, and for some people quite a different approach to innovation - although personally I think you could apply it to any field.


So start with a problem, scope out several potential solutions, and then roughly work out what the risks in each of those are and how you are going to mitigate them - often you are truing to work out which is the "least worst". Taking SPADs: the problem is a signal not being seen, potential solutions include driver training, driver shift reduction, change of design of the signal, in cab signalling, radio link from the signal to the train, track induction link from the track to the train, double signals (the old home and distant), vision system (that's just what I came up with while writing that sentence, I'm not pretending it's an exhaustive list). Then for each of those there are a range of risks (safety and reliability), costs and benefits, and what you'll probably come up with is a solution that looks pretty much like ATP funnily enough! Also remember that the simpler and lower tech a solution is a) the more reliable it is likely to be and b) the easier it is to prove it's safe (because it has less failure modes).


So if you're promoting an innovation for a safety critical environment, somehow you need to get it in the minds of the right people when they are going through that selection of options stage, which is a HUGE challenge for the industry. But you must have an argument ready for every immediately apparent safety and reliability issue - which for vision systems unfortunately are many and varied.


For anyone interested in understanding this process further, a good starting point is RSSB's "Taking Safe Decisions" www.rssb.co.uk/.../2014-guidance-taking-safe-decisions.pdf


Don't know if I've explained that very well, very happy to discuss and clarify further! Once you move away form a particular case and get into the more general case it becomes very interesting - and quite important - which I suppose is why several universities are now running master's degrees in this stuff!


Cheers,


Andy


 

Hi James,


Apologies for my slow reply - I got distracted into other less useful discussions on this forum! (Literally don't go there!)


The problem is the very high levels of safety and reliability you need. Even if you say this is an add-on that supplements rather than replaces the driver, so that the driver is the main control of the brakes and this has additional control, there are two huge problems. One is that it must not apply the brakes when it shouldn't - emergency braking a train is considered a hazard (there have been a number of serious passenger injuries caused by it). So it must never falsly think that something is on the track when it isn't. The other is more subtle, and was found in early adoptions of Automatic Train Protection. Once the driver knows there is an automatic protection system in place they will tend to rely on it, and if it isn't fail-safe horrible things can happen... These may be solveable, but are they solveable at a cost that is less than ETCS?


What may be interesting is a simple alarm system, so if a visual system picks up what it thinks is a hazard it alerts the driver who then decides what to do about it. That would have far fewer safety implications. But given the cost of railway approvals, due to the fact that no possible risk must come from any change, it would be 'interesting' getting it supported by major authorities. (Ref my comments above about this not influencing the accidents we're actually seeing.)  If it were me, I'd start by looking at it as a tramway solution - which I'll bet someone already is!


You're probably aware that SPADs are a pure financial issue, the UK ATP system is in place for this, just needs to be rolled out more. (Or ETCS.) SPADs aren't quite as scary as the "red-tops" liked to make out in the early 2000s, there are huge protection zones in place. But they're not good of course - every time you degrade a safety margin then by definition there's a bit less left!


But regarding driverless trains generally, no problem, we've got them. And they meet the levels of safety demanded. If only we could afford them...


Rule one of railway engineering: Nothing is cheap, nothing is simple, unless you can persuade the public to compromise on safety,

Rule two of railway (and indeed safety) engineering: Every time you eliminate one Hazard, you create ten more!



It is very, very frustrating, being someone who is by instinct an innovation manager, to have to write posts like this. And as you can imagine, since I moved to the rail industry I've had many, many discussions like these. We desperately need innovations, particularly low cost innovations. But my (long!) experience has been that the only approach I've seen work is to start by working out how you're going to do it (whatever "it" is) safely and reliably at a "broad brush" level, and then work out the detailed technical solution - which is much the easier bit. Trying to start with a piece of whizzy technology and then trying to make it safe is a nightmare, indeed almost impossible. It's a strange business, and for some people quite a different approach to innovation - although personally I think you could apply it to any field.


So start with a problem, scope out several potential solutions, and then roughly work out what the risks in each of those are and how you are going to mitigate them - often you are truing to work out which is the "least worst". Taking SPADs: the problem is a signal not being seen, potential solutions include driver training, driver shift reduction, change of design of the signal, in cab signalling, radio link from the signal to the train, track induction link from the track to the train, double signals (the old home and distant), vision system (that's just what I came up with while writing that sentence, I'm not pretending it's an exhaustive list). Then for each of those there are a range of risks (safety and reliability), costs and benefits, and what you'll probably come up with is a solution that looks pretty much like ATP funnily enough! Also remember that the simpler and lower tech a solution is a) the more reliable it is likely to be and b) the easier it is to prove it's safe (because it has less failure modes).


So if you're promoting an innovation for a safety critical environment, somehow you need to get it in the minds of the right people when they are going through that selection of options stage, which is a HUGE challenge for the industry. But you must have an argument ready for every immediately apparent safety and reliability issue - which for vision systems unfortunately are many and varied.


For anyone interested in understanding this process further, a good starting point is RSSB's "Taking Safe Decisions" www.rssb.co.uk/.../2014-guidance-taking-safe-decisions.pdf


Don't know if I've explained that very well, very happy to discuss and clarify further! Once you move away form a particular case and get into the more general case it becomes very interesting - and quite important - which I suppose is why several universities are now running master's degrees in this stuff!


Cheers,


Andy