Wednesday, September 22, 2010

3000 grands prix per second


OFFICIAL PRESS RELEASE

Woking, United Kingdom, Sep 22, 2010

It�s lap seven of the Singapore Grand Prix. Your driver�s pace on the Option tyre has started to look a bit worrying, and he�s stuck behind an unusually out-of-position slower car. Suddenly, a driver loses the back-end at Turn 17 and a Safety Car is quickly deployed to clear the debris.



You�ve got 20 seconds to make a decision: box your driver this lap to change his tyres, or keep him out behind the Safety Car? It�s a call that could win or lose you the race.

Feeling the pressure yet?

For James Miskin, operational performance team leader at Vodafone McLaren Mercedes, this isn�t an unusual situation. In fact, James and his team will doubtless have rehearsed this exact scenario many times before.

About five million times, to be exact�

How so? Because James� three-man strategy team relies on phenomenal computing power to crunch through millions upon millions of likely scenarios before each and every grand prix.

James explains: �After qualifying, we carry out around five million experiments, looking at how the race could pan out for Jenson and Lewis. For instance, it takes about an hour to model all the possible Safety Car outcomes.�

Clearly, when it comes to sheer muscle, the Vodafone McLaren Mercedes strategy department is seriously indebted to its computing power: �Last year, we replaced six PCs with three new Lenovo ThinkStation D20 machines. We�ve gone from running around 600 race simulations per second up to around 3000.

�That�s a five-time improvement, and it means we finish all our post-qualifying simulations two hours after the session�s finished, rather than having to wait until Sunday morning. That makes a massive difference.

�And we need to run a lot of different scenarios. For example, what if the Option tyre degrades faster than we initially thought? What if the Option is a lot quicker than the Prime? What if the time lost in the pitlane is worse than we�d anticipated?

�Actual driver laptime doesn�t matter � it�s the differences that are important, and whether we�re faster or slower than other people. The laptime only affects how long the race is. So we�ll pick a few drivers of differing paces � for instance, both Jenson and Lewis, a couple more in the midfield, and a quicker driver. What should the quickest driver do? What should a Force India that sneaks into Q3 do? What should a fast guy who doesn�t make Q3 do?

�In the race simulation model, we have around 5000 different parameters, and we try a range of different strategies for each driver: starting on Prime, starting on Option, a differing range of stints. And that gives you an idea of where the field should be stopping, who�ll start triggering the pitstops, what the risk is if you don�t trigger them and somebody else goes first.

�The ability to accurately and rapidly simulate the expected outcomes of a complicated series of events in a grand prix is a very valuable tool,� concludes Miskin. � It gives us the power, flexibility and foresight to deal with almost any eventuality � and it�s that ability to be flexible, and think on your feet, that can win or lose you a grand prix.�

* Official details courtesy of VODAFONE MCLAREN MERCEDES *

Copyright � 2010, car-evolutioncars. All rights reserved.

No comments:

Post a Comment