Sunday, November 19, 2017

How to overcome problems with the MRT signal system

SMRT faced a lot of problems in introducing a new signalling system. It has led to many breakdowns and one serious collision due to "signal problems".

I asked the question - what are the signal problems. A few people explained that SMRT is trying to run two signal systems. The train has to change from one signal system to another system on the journey.

This is a bad way to implement the change. If I were consulted, I would never recommend this approach. It is based on my "common sense".

The experts may give some "technical reasons" for adopting their approach. They will find it very hard to convince me.

What is my approach?

Take a look at Google trying to implement a driverless car on the road. The technical problems that Google has to overcome is porbably ten times more complex that a driverless train running on a track.

How does Google approach it?

They develop a driverless car using sensors. The sensor will look for other driverless car on the road, but will also allow for manually driven cars and pedestrians and even dogs running on the road.

The driverless cars detect these objects and other driverless cars on the road.

If one lane contains only driverless cars, it can give the instruction for all the cars to travel at higher speed. If any driverless car has to slow down to change lane or avoid a "foreign object", the other cars get the signal and will slow down together.

The same approach can be used for the train system. It should cater for driverless train and mannually driven trains. It should also cater for people or dogs falling on to the track.

If all the trains are driverless, it is possible to get them to travel at higher speed. If some trains are manually driven, they will be travelling at a slower speed. The other trains will adjust their speed accordingly.

All the trains need to have sensors to detech "foreign objects" and to transmit their speed to the control room. The control room can then set the maximum speed that the train can travel.

If there is a stationary object ahead, the train has to stop. The speed of braking depends on the distance.

Under this concept, there is no need for any train to "change from one signal system to another system".

The AI in each train will take the data from its sensors, e.g. to look at "foreign objects" and use it to slow down when necessary. The speed of the train is sent to the control room.

The control room knows the speed of all the trains on the tracks and can indicate the speed that the train can travel. This speed has to be reduced, if necessary, based on any obstructions that the actual train may sense on the road.

Under my system, I would also expert some other sensors at the station and on the tracks to send signals of foreign objects on the track and whether these objects are stationary or moving. These signals can be processed in the control algorithm.

The ideal operating condition is that all the trains on the line are driverless and there is no foreign object. The trains can all travel fast and achieve the goal of 1 train arriving every 2 minutes at the station. We will reach this stage eventually when all the trains have been converted to the driverless system.

I may not be an expert in train control systems, but I have some "common sense". If any experts wishes to add their views to help the discussion, they are welcomed to do so.

Tan Kin Lian

No comments:

Blog Archive