Humans look to robot race

One thing that cars do remarkably well is kill people. Each year in the UK around three and a half thousand road users and pedestrians are killed -- and a hundred times as many injured. While there has been a steady stream of innovations designed to reduce the chances of a collision hurting a vehicle's occupants -- seatbelts, airbags, crumple zones -- there's been little movement on the much smarter business of stopping collisions happening at all. Even where pure legislation can effect changes, as in acceptable alcohol and speed limits, nothing much changes from one decade to the next. The most important recent safety innovation has been the hyperbright LED brakelight, which turns on instantly and adds a few hundred milliseconds extra braking time for the car behind. Worth having, but hardly sufficient.

Wasn't the future supposed to be different? Automatic road systems were promised, with robotically steered cars forming up into tight convoys and screaming across the countryside. Radar-guided collision avoidance systems would take care of safety far better than us fallible humans, while central routing would avoid congestion and billions of pounds' worth of wasted fuel and time. Fantastic.

If only it had happened that way. We're left with a few ancient TV reports from Volkswagen's test track, and cars that bleep quietly when you're about to back into a lamp-post. Driving a vehicle needs a great deal of intelligence or, in the case of motorcycle couriers, the unshakeable belief that you are on a mission from God and protected by all His great angels from harm. As modern computers have yet to evince real smarts or a theological bent, it's still us all-too-mortals behind the wheel. The commercial imperative to do enough research to fix this is missing.

There is another way. The early days of aviation were marked by similar research longeurs, as the uncertain finances of the fledgling industry left basic research and development undone. Various individuals, seeing this as an abdication of patriotic duty -- patriotism being more important than economics back then, with some splendid wars to prove it -- decided to force the issue by running and paying for flying competitions. Others then stumped up the cash to enter, with companies and nations vying for records. The Spitfire, that icon of British fortitude, came about like this in an amazing story involving fascist showgirls, but that's a story for another time.

Now, seventy years later but hopefully without the fascist showgirls, the Pentagon is doing the same for unmanned road vehicles. The US defence agency DARPA has put up a million dollars for the winner of a Grand Challenge race to be held on 13 March, 2004, in the Mojave Desert between Los Angeles and Las Vegas. The 250-odd mile course will have to be completed at an average speed of above twenty miles an hour by the autonomous robot cars -- and just to add spice to proceedings, the details of the route will only be revealed a couple of hours before the off. The parallels with Fear and Loathing are too strong to resist, and I do not intend to try.

Fantastic stuff. Teams are already registering, with some 25 groups already listed on the Web site. The discussion forums are buzzing with lots of good questions being asked: do you convert an existing vehicle or start from scratch? Can you relay pictures back from each contestant? But, ominously, the biggest, longest-lived and most ferociously debated thread is one about the rules.

Despite the many good and exciting engineering challenges, the thing that threatens to sink the entire endeavour and turn it from a noble exercise in generating battle robots to a live version of Wacky Races is the rules. At one point, DARPA set a six-hour time limit, which as was quickly pointed out by the competitors was undoable by a human driver. And you can imagine the fun with insurance. DARPA won't guarantee that the race course will be kept clear of other vehicles or people, and mandates electronic emergency stop systems and chase vehicles, but there are some odd things out there in the desert. The million dollar prize money is taxable, and the investment for an entry will probably pan out at half a million at least -- not such a sweet deal.

But perhaps the most restrictive rule is that only American teams can enter. That may seem to make sense -- after all, it's the American military paying American tax dollars in return for American know-how -- but in practice means that the competition will be less fierce, the ideas less ambitious and the results much less useful. The grand race organisers of the 20s and 30s knew the importance of fighting your neighbours, not your family.

So the DARPA Grand Challenge is going to be a fraction as interesting and useful as it could be. The world needs much smarter cars designed not to kill people much more than it does robot jeeps -- and in any case these days, spinoffs go from commercial to military rather than vice-versa. It's worth watching, but a race that's just that little bit smarter would go a lot further for all concerned.