3 ms·

> The quickest it will ever slow down is with its emergency brakes. What about derailments? Can a derailment lead to a higher deceleration than the emergency b

> The quickest it will ever slow down is with its emergency brakes. What about derailments? Can a derailment lead to a higher deceleration than the emergency brakes are able to produce? (I don’t know the answer, genuinely curious.) > So the minimum headway is the reaction time (milliseconds if automatic) How do the following train detects that the leader train is emergency breaking? Is it based on trackside sensors, or some train-to-train communication?

londons_explore 4y ago
> Can a derailment lead to a higher deceleration than the emergency brakes are able to produce? Yes. But a derailment of one carriage of a 5 carriage train will already likely cause the derailment of the other carriages, and in my proposal the carriages of the train behind. Derailments at speed are already typically deadly. > How do the following train detects that the leader train is emergency breaking? I would use train 2 train communication to say "I am not emergency braking" every 1 millisecond. If more than a few of these messages aren't received, then you have to assume the train in front is emergency braking, and act accordingly. A design like that is important because infrastructure failures are frequently coincident - eg. power loss causes signalling and comms failure due to an untested UPS.
- londons_explore 4y ago
  In a more detailed design, you would have every train publish a "movement plan" every few milliseconds. The plan contains the current location (with uncertainty bounds), and future locations. The plan also contains alternate future locations that could happen in various emergency or failure situations - and one of those is to apply the emergency brakes. A system checks each plan against the plans of all other trains, and approves them if they are non-colliding. The 'alternate' plans can overlap only if to do so indicates independant failures must have occurred - for example the brakes of train 1 fail at the same time as the traction motor of train 2 gets stuck on max, and comms fail. Track infrastructure such as signals and points would also submit plans - so that for example a set of points can prevent any trains passing over while it isn't in a safe locked state. All trains must then follow one of the plans they have approved. If comms fail and they can't get a new plan approved, they must still follow an approved plan - which means they are guaranteed not to collide with any other train. All plans would be signed with timestamps (to prevent someone maliciously changing them to cause an accident), and would normally be delivered by dedicated rail infrastructure, but could be delivered by 3g/4g/wifi in an emergency. Manual plans could also be entered and distributed, for example via USB stick, for maintenance - such plans would say "I have access to this huge segment of track, and nobody else may enter".
  - heavenlyblue 4y ago
    How would you “plan” for a brake failure?
    - londons_explore 4y ago
      So your train publishes say 3 plans: Plan A is used if all systems remain operational. It says that the train will continue to move at 60 +- 5 mph for the next 10 seconds, then slow gradually to 45 mph. The plan could consist of a section of "reserved" track for each second in the future, and the train is sticking to the plan if it remains within the reservation. Plan B is used if service brake failure happens. It says the train will apply its emergency brake and slow to zero at a given rate. Plan C is used if both the service brake and the emergency brake fail. It says the train will continue at 60 mph, and may speed up or slow down a little, before finally guaranteeing to stop within 30 miles. All the plans would be written automatically by software, and each would be 'conservative' - ie. if there are any uncertainties, the plan must include them. Less advanced trains would have wider bigger plans, and must therefore keep wider separation distances. Plan C is obviously very unlikely - and is therefore allowed to overlap with similar plans B and C of other trains - the reasoning being it is very unlikely for the service and emergency brakes of this train to fail, at the same exact time as the emergency and service brakes of another train.
- krisoft 4y ago
  > I would use train 2 train communication to say "I am not emergency braking" every 1 millisecond. I like that. How would that work when there is no lead train? In other words how would a train know that it is not hearing from the train before it because there is no train, vs not hearing because the lead train is damaged/emergency braking.
  - londons_explore 4y ago
    See my other response in this thread - an actual deployment would require every train on the tracks to automatically submit a 'plan' for it's location, and if it cannot submit a plan, then it isn't allowed on the tracks. Plans last until replaced by a new plan, so a train will always have at least a place on the track it can be. It doesn't solve the problem of 'there is a tree on the tracks' or 'someone put a homemade train on the tracks' - you need a different solution to that.
- bombcar 4y ago
  Right now a derailment brings down one train (perhaps deadly). Your proposal would have the derailment bring down that train, and have another train potentially slam into it. Derailments stop a train much faster than the brakes ever could. https://www.seattletimes.com/seattle-news/train-derails-from-bridge-onto-interstate-5-near-olympia/ https://www.seattletimes.com/seattle-news/train-derails-from...

londons_explore 4y ago

> Can a derailment lead to a higher deceleration than the emergency brakes are able to produce? Yes. But a derailment of one carriage of a 5 carriage train will already likely cause the derailment of the other carriages, and in my proposal the carriages of the train behind. Derailments at speed are already typically deadly. > How do the following train detects that the leader train is emergency breaking? I would use train 2 train communication to say "I am not emergency braking" every 1 millisecond. If more than a few of these messages aren't received, then you have to assume the train in front is emergency braking, and act accordingly. A design like that is important because infrastructure failures are frequently coincident - eg. power loss causes signalling and comms failure due to an untested UPS.

londons_explore 4y ago
In a more detailed design, you would have every train publish a "movement plan" every few milliseconds. The plan contains the current location (with uncertainty bounds), and future locations. The plan also contains alternate future locations that could happen in various emergency or failure situations - and one of those is to apply the emergency brakes. A system checks each plan against the plans of all other trains, and approves them if they are non-colliding. The 'alternate' plans can overlap only if to do so indicates independant failures must have occurred - for example the brakes of train 1 fail at the same time as the traction motor of train 2 gets stuck on max, and comms fail. Track infrastructure such as signals and points would also submit plans - so that for example a set of points can prevent any trains passing over while it isn't in a safe locked state. All trains must then follow one of the plans they have approved. If comms fail and they can't get a new plan approved, they must still follow an approved plan - which means they are guaranteed not to collide with any other train. All plans would be signed with timestamps (to prevent someone maliciously changing them to cause an accident), and would normally be delivered by dedicated rail infrastructure, but could be delivered by 3g/4g/wifi in an emergency. Manual plans could also be entered and distributed, for example via USB stick, for maintenance - such plans would say "I have access to this huge segment of track, and nobody else may enter".
- heavenlyblue 4y ago
  How would you “plan” for a brake failure?
  - londons_explore 4y ago
    So your train publishes say 3 plans: Plan A is used if all systems remain operational. It says that the train will continue to move at 60 +- 5 mph for the next 10 seconds, then slow gradually to 45 mph. The plan could consist of a section of "reserved" track for each second in the future, and the train is sticking to the plan if it remains within the reservation. Plan B is used if service brake failure happens. It says the train will apply its emergency brake and slow to zero at a given rate. Plan C is used if both the service brake and the emergency brake fail. It says the train will continue at 60 mph, and may speed up or slow down a little, before finally guaranteeing to stop within 30 miles. All the plans would be written automatically by software, and each would be 'conservative' - ie. if there are any uncertainties, the plan must include them. Less advanced trains would have wider bigger plans, and must therefore keep wider separation distances. Plan C is obviously very unlikely - and is therefore allowed to overlap with similar plans B and C of other trains - the reasoning being it is very unlikely for the service and emergency brakes of this train to fail, at the same exact time as the emergency and service brakes of another train.
krisoft 4y ago
> I would use train 2 train communication to say "I am not emergency braking" every 1 millisecond. I like that. How would that work when there is no lead train? In other words how would a train know that it is not hearing from the train before it because there is no train, vs not hearing because the lead train is damaged/emergency braking.
- londons_explore 4y ago
  See my other response in this thread - an actual deployment would require every train on the tracks to automatically submit a 'plan' for it's location, and if it cannot submit a plan, then it isn't allowed on the tracks. Plans last until replaced by a new plan, so a train will always have at least a place on the track it can be. It doesn't solve the problem of 'there is a tree on the tracks' or 'someone put a homemade train on the tracks' - you need a different solution to that.
bombcar 4y ago
Right now a derailment brings down one train (perhaps deadly). Your proposal would have the derailment bring down that train, and have another train potentially slam into it. Derailments stop a train much faster than the brakes ever could. https://www.seattletimes.com/seattle-news/train-derails-from-bridge-onto-interstate-5-near-olympia/ https://www.seattletimes.com/seattle-news/train-derails-from...