Blog

Why self-driving cars could soon be much more affordable

16th November 2016
Anna Flockett
0

People were understandably sceptical when George Hotz, an artificial intelligence software developer and renowned hacker, announced the $999 Comma One kit, which promised to convert normal cars into semi-autonomous ones. OK, so the kit had its limitations, which Hotz’s company was at pains to point out. It would still have needed the driver to pay close attention at all times. 

Availability was to be limited to a handful of car models, and users would have needed to pay a monthly $24 subscription on top of the unit price.

A letter from the USA’s National Highway Traffic Safety Administration, highlighting concerns about road users’ safety, saw Hotz’s company cancel the self-driving car project in October. However, the initiative did serve to underline a change in the way we think about autonomous and semi-autonomous cars: yes, they’re expensive at the moment, but that won’t always be the case.

Early pioneers of advanced, self-driving features have put these into their pricey, top-end models: the Mercedes S-Class costs in excess of $100,000, while Tesla’s Model S will set you back upwards of $66,000. Google’s self-driving cars, which aren’t on sale, contain around $150,000-worth of equipment.

One of the main reasons autonomous vehicles are so pricey is that they contain a lot of costly components, notably all the complex sensors you need to keep tabs on the road and other vehicles. But the bill of materials is also where you can achieve the biggest savings, thanks to the well-trodden path where innovation is followed by competition and economies of scale. Prices of pioneering tech start off high and then come down, as the overall market expands and manufacturers refine the way they work.

And even though we could be waiting a long time before fully self-driving cars become commonplace, prices of the components needed to build them are already falling. This is because there are already other markets for some of them: ultrasonic sensors, radar and cameras are required to enable existing features such as parking assist, smart cruise control and lane-keeping.

Prices are even coming down on components where there isn’t currently a market. Because the future market for self-driving vehicles is anticipated to be so large, manufacturers have been willing to invest in innovation, in an attempt to get their parts into early models. According to forecasters, the autonomous and semi-autonomous vehicle market could be worth more than $40BN by 2025. Sensors are expected to account for a third of this.

Lidar – laser-based radar that creates a 3D map of its surroundings using near-infrared light – is a perfect example of how market forces encourage innovation and drive down prices, even when there isn’t yet a market for something. Rooftop-mounted domes or masts that house Lidar scanners, such as those found on Google’s self-driving cars, have quickly become the defining feature in popular images of autonomous vehicles.

Self-driving cars require a combination of sensor types. Radar penetrates through fog and rain, but doesn’t offer a great deal of precision. Optical image sensors enable the vehicle to see what the human eye can see. Ultrasonic sensors, while their range is limited, are ideal when it comes to detecting things close to the car. Lidar, meanwhile, can instantly measure distances, even in challenging light conditions where the human eye or an optical sensor would struggle.

The majority of Lidar scanners work via a beam that rotates incredibly quickly. This is why the unit has to be mounted on the roof of Google’s self-driving cars, as it’s the only place where you can get an all-round view without obstructions. The fact that Lidar scanners typically come with a price tag as high as $75,000 shows why most autonomous vehicle pioneers have used just one such sensor in their designs.

Traditional Lidar buyers, including those working in the aerospace industry, have accepted high prices, partly because they typically only buy small volumes. The automotive sector is a different proposition. A much bigger market for Lidar sensors – potentially tens of millions of units per year compared to several thousand in the aerospace world – has appeared, but manufacturers want lower prices. It’s forced those involved in sensor manufacturing to come up with new ideas.

As a result, you can now buy a rotating Lidar scanner for less than $10,000. What’s more, 2017 is expected to see solid state optical phased array Lidar units come to market for just $250. While solid state Lidar doesn’t give the all-round view that traditional Lidar can, by incorporating three or four such sensors into the car body, you can rid yourself of the dome or mast on the roof – and still come in at a fraction of the cost.

Of course, the elephant in the room for all those working to deliver lower-priced Lidar scanners to the self-driving car market, is that Tesla, the leading autonomous car maker, doesn’t use Lidar.

Tesla claims that radar, cameras and advanced software can be used achieve what Lidar does. The belief in radar’s superiority over Lidar is a sentiment CEO Elon Musk has reiterated since the fatal accident in May, when a Tesla Model S and its driver didn’t notice a white trailer against a brightly lit sky up ahead. Tragically, a Lidar sensor may have helped in this situation.

The terrible accident underlines the risks of developing self-driving vehicles, and the challenges that pioneers such as Musk and Hotz face when it comes to bringing this technology to the mainstream market. It also demonstrates how falling prices of Lidar sensors and other autonomous car parts could help improve safety, at the same time as making self-driving vehicles of the future more budget-friendly than today’s.

By Mark Patrick, Mouser Electronics 

Featured products

Product Spotlight

Upcoming Events

View all events
Newsletter
Latest global electronics news
© Copyright 2024 Electronic Specifier