Elon Musk has become something of an internet icon over the last year. From sending a Tesla Roadster into space to building and selling a literal flamethrower (that’s not actually a flamethrower, apparently), he’s used to doing the things everyone else would be scared to try. It makes sense, then, that he’s moving into electric cars.
And he’s not being shy about it, either – predicting that the technology will turn Tesla into a company worth $500 billion. That’s a bold prediction, especially when coupled with a rejection of two of the main elements of everyone else’s attempts at making driverless cars. He recently called Lidar and HD maps “the two main crutches that should not be used”, as well as branding them “obviously false and foolish”. This has caused considerable annoyance amongst the people that build
both these things into self-driving cars for a living – but the average consumer is probably a little stumped.
Fear not – in this post, we’re going to explain what exactly each of those things are, why they’re so important to electric cars, and how Tesla not using them could make a difference in how their self-driving cars of the future end up operating.
What is LIDAR?
Lidar is a specific type of sensor that uses lasers to calculate the distance between it and other objects. By measuring the time it takes for the laser to bounce back off an object, a driverless car with LIDAR can figure out what obstacles are around it and work to avoid them.
What is HD mapping?
HD mapping, as the name implies, involves the creation of extremely high-definition maps, accurate down to centimetres. Including things like the height of curves in a road, these maps work to help the driverless car understand its surroundings, allowing it to make decisions that humans take for granted, like waiting for a traffic light to change colour. Having the maps available also means the cars can do this with a smaller time delay, leading to more seamless movement.
An example of HD mapping in action…
Why does Elon Musk dislike both of them?
Musk’s main issue with both these technologies is the problems it can run into when its surroundings change before they do. Accuracy is, of course, key to the running of driverless cars, and if inaccuracy leads to failure, less people will invest in driverless cars because of it.
But these cars still need to navigate the city somehow – so what is Tesla putting in its place? The answer is deceptively simple: cameras. Each of Tesla’s driverless vehicles are equipped with lots of them, covering all angles with the aim of teaching the car to “see” what’s around them. A forward-facing radar and ultrasonic sensors placed around the vehicle add to the car’s potential field of vision. The Financial Times also notes its reveal two weeks ago of a computer chip, designed in-house that will be installed in all its vehicles. The specs themselves are mind-blowing, but the chip’s real significance is that, according to Musk, it can handle all the data input from these sensors and cameras more effectively than any of the other chips currently on the market.
Tesla also has some complex software onboard to offset the chances of things going wrong. Much of that is grounded in AI, or artificial intelligence, specifically “neural networks”. In essence, these are
networks that simulate the structure of the human brain – with its billions of connected cells – that allow humanlike decisions to be made. Like other AI, the network can learn by itself, but does need
to be programmed first. Tesla’s way of doing this is to feed the network lots and lots of data about the objects that the car might encounter – data that’s currently being captured by the 40,000 standard Tesla cars already on the road.
All this might make for a smoother driving experience eventually, but as with any new technology, these neural networks do have their own problems. Neural networks can go slightly haywire when they see an object that doesn’t match what they have in their data banks – and with the world always changing, these networks might not be able to keep up with demand. Just look at some of the experiences of the lucky attendees that got to demo the “AutoPilot” system at Tesla’s Autonomy
Day: Trip Chowdhry, managing director at Global Equities Research, noted to Wired that it didn’t recognise a turn arrow on a traffic light. Though the driving was smooth, these issues make Tesla’s ambitious time scale hard to believe: will they really have fully driverless cars by the end of 2019?
Well, Tesla is notoriously flexible on its deadlines, so that may not come to pass. But Elon Musk and co are looking to reinvent what we think we know about driverless cars – and if they do it as well as
they say they can, waiting might not matter.
What are your thoughts on Tesla’s ambitions for their driverless cars? Do you think disowning all these common technologies is a good or a bad move? Let us know in the comments!