The Sandbox: the AVSandbox knowledge hub

The has-been-trieds of technology. They’re inspiring, they’re disheartening. Those innovators who like to chase firsts and ride the moment are in a complicated relationship with them. So what feelings does it stir, when we remind ourselves that the hotly contested self-driving car was already being successfully driven around as far back as the 80s? Before the advent of electronics, brilliant and determined minds made do with what they could. Cruise control was invented all the way in 1948, and in fact, Da Vinci himself built a self-propelling cart, sometimes considered the first robot, that was powered by a spring and could change trajectory from a pre-set mechanism. But the story of the ‘modern’ self-driving cars really started in 1977 and quickly turned into something that defied reality.

This was in Japan, where the Tsukuba Mechanical Engineering Lab built a rail-car that controlled itself by reading street signs – using early and highly limited forms of computer vision. It could drive at speeds of up to 33km/h. In the wake of this success, the world became ready to make a leap-of-faith.

  In just 10 years, in 1987, the European government launched the great Prometheus project, the largest single research grant that’s still ever been given to autonomous vehicles. €750 million was going to study avenues of tech that we’re still studying now – from cockpit driver assistance to intervehicle coordination. The hope was to find ways to make traffic safer and smoother. It was first anticipated that part-car and part-infrastructure solutions would be necessary, and the idea was to have cables buried under roads to aid navigation. But the end result surpassed all expectations.

It was all thanks to an engineer by the name of Ernest Dickmanns. Maturing his career during the space age, he became interested in computer vision for satellites observing the earth, but took his work into driverless cars. And so in response to the project, he built the groundbreaking VAMORS machine. It was a 5-ton Mercedes carrying a burdensome computer, equipped with its own senses and eyes. By studying areas of high contrast and colour changes captured by the cameras, the car did something truly impressive: it rode the German autobahn, and at 96km/h. Working with basic computer vision that had limited power and speed, he and his colleagues invented algorithms that could extract information about the environment using as little photographic data as possible. In the spirit of true, gallant engineering, they also decided that they’d be the car’s passengers, to force themselves to design it safely.

By 1994, the project made even more remarkable achievements. Now in the form of the cars VAMP and VITA-2, and with more cameras, better noise filtering, and a way for the cameras to point themselves to points of interest on the road, the vehicles were let loose in the heart of busy driving: Paris. They navigated 3-lane roads, recognizing other cars and performing safe overtaking manoeuvres. Clocking 1000km of distance there, with speeds of up to 131km/h, they moved onto a road trip between Germany and Denmark. There, they drove 2000km with a top speed of 180km/h. Each maneuver had to be approved by the supervising driver, but there was an average of 9km between each human intervention, and even some 150km of uninterrupted autonomous driving. So what emotions does this invoke? Those who are jealous about sharing glory can feel some relief. As amazing as these achievements are, they barely scratch the surface of what’s needed for scalable autonomous vehicles. For one, every move the VAMORS, VAMP and VITA performed needs to be engineered to consistent perfection, and two, instead of struggling to collect data on the surrounding environment, we struggle to fuse the sheer vastness of it correctly. Some problems still remain: the Prometheus vehicles struggled with negative obstacles like potholes and puddles, and that hasn’t been conclusively solved. Perhaps we are defined by a different era of self-driving cars – the one based on the technology showcased in the DARPA competitions of the 2000s, and the promises of machine learning. But the story of this field is filled with truly ingenious and bold engineering, and we can only hope to do it justice. 

And now, a little about us…

We’re Claytex, self-driving car safety experts. Instead of driving AVs for millions of miles, waiting for dangerous road incidents to happen, we want to anticipate them all in massive parallel simulations that can comprehensively test how safe the car is. If you found this article interesting and want to learn more, please get in touch.

Written by: David Dubinsky – Project Engineer

Please get in touch if you have any questions or have got a topic in mind that you would like us to write about. You can submit your questions / topics via: Tech Blog Questions / Topic Suggestion