December 01, 2010

How to Increase Speed of Trains in India

The Basic Philosophy

The solution to any real life problem consists of two components -- a hard technology component and a soft human component. While speeds of trains can of course be increased with more powerful engines, lighter materials, better aerodynamics and intelligent signalling, what we also need is a way to ensure that everyone in the railway system has a vested interest in ensuring that speeds and schedules are maintained,  is motivated enough to go the extra mile to make that happen and has the information at his or her finger tips to act on the motivation. This is the goal of TERM -- the Technology Enabled Real-time Motivation initiative.

The Core  Idea

Who is responsible for delays ? At the highest level, the Railway Board ! But we need to establish a more granular level of responsibility. Is it the signal operator ? the station master ? the staff of a particular station ? the staff of a particular group of stations ? of a specific division ? of a zone ? Or is it some other specific collective of individuals ?

SIMRAN is GPS based system that the railways plan to use to track the position of all trains at all times. [ see 1, 2, and presentation by Dhande of IIT Kanpur ] TERM will use the data available from SIMRAN to create a “delay model” within the Indian Railway system. When water flows through a network of pipes, any blockage anywhere causes a  buildup of pressure in certain segments of the network. Similarly, the delay model will create real-time image of the delay hot-spots, the “pressure” points, within the system by adding up the instantaneous individual delay of all trains in a defined area.  By integrating  this instantaneous total delay data of all trains in the area over a period of time we will arrive at a cumulative delay-index for a “collective” ( which could be station, region, zone or some other defined collective ) and this will be used to determine a financial incentive for everyone in that collective. This recognises that individuals cannot be held individually responsible for delays but collectively are  indeed responsible and so everyone in the collective should be incentivised to minimise the delay-index. Individuals could belong to overlapping collectives and  individual incentives could be also calculated depending on the sophistication of the model.

If a collection of individuals are to be held accountable for delays, then it is only fair that they be individually informed of delays as they get accumulated in real-time and this is the second component of TERM.  SIMRAN will be generating megabytes of position/delay data every minute and we need a system that will rapidly and in real-time scan the data -- using a Map-Reduce algorithm on a Hadoop platform that will constantly monitor this data flood -- to isolate and prioritise the most important hot-spots. This is roughly analogous to the Google Page-Rank mechanism that dynamically identifies top pages or trending topics of the the moment.

Individuals in each collective  would have a graded hierarchy of responsibility and depending on the quantum of delay at any given time, SMS messages will be sent to specific individuals alerting them to the delay that is building up in the area of his collective. For example a 1 minute will alert the signalman, a 5 minute delay will alert the station master, a 10 minute delay will alert the Divisional Manager and so on upto the General Manager, who will also be a  part of the collective incentive scheme. The actual escalation matrix will have to be determined based specific authorities and responsibilities and would change dynamically based on actual duty hours. Net-net no one will be able to absolve themselves from the responsibility for the delay-index accumulated in his collective.

The Tentative Architecture



The Bottom Line

TERM will use existing GPS data from SIMRAN to create unique mechanism where everyone in the system is encouraged to minimise delay and earn financial incentives for the same.