How To Make A Regular Car Self-Driving In Three Days

Currently the names we most strongly associate with autonomous vehicle technology—Tesla, Google, Uber, perhaps other legacy automakers like Audi or Volvo—are large companies, able to allocate huge economic resources and legions of employees to the task.

But as exposure to autonomous systems grows more common in both civil society and the technology world, space is opening up for smaller, scrappier players who can build on the groundwork laid by pioneers to turn prototypes into products in a short space of time, and at far lower cost than would have been possible just a few years ago.

One such company is Voyage, a Silicon Valley startup with a stated end goal of building a fleet of self-driving taxis, but which is also willing to share details of its R&D process along the way. Recently Voyage gave a behind-the-scenes look at its most recent project, where a small team of engineers retrofitted a factory standard Ford Fusion to drive in autonomous mode.

The process was carried out over three days at APROE, a prototype engineering firm in San Francisco which provides workshop space and the ability to manufacture custom parts on-site.

The team first disassembled the front and rear bumpers of the car in order to fit long- and short-range radar units, some of which were attached to a bespoke mount to allow for greater customization further down the line (since there's no standard size for radar sensors at present).

Next up they built a frame to fix the Velodyne LIDAR sensor to the roof, another task that required custom machining since the unit is rare enough that there are no off-the-shelf mounts available.

With all the sensors in place, inputs and outputs were connected to a readymade "drive by wire" kit from Dataspeed, a company that specializes in making control systems for vehicles and industrial robots.

The Dataspeed kit was crucial to the project, Voyage CEO Oliver Cameron explained in a call, because it has already been configured to work with the Ford Fusion's CAN bus: the protocol which lets the different microcontrollers in a modern vehicle talk to each other.

"Each of the commands on the CAN bus are different per vehicle, so if you set out to 'hack' a car, the work you do on one car will not transfer to another car," Cameron said.

This meant that by choosing to work with the Dataspeed system and a Ford Fusion, the team was spared the complexity of trying to reverse engineer the car's electronics, letting them focus on getting the autonomous control system up and running.

By cutting out unnecessary work Voyage hopes to spend more time on their core product, which Cameron stresses is not the vehicle itself (which will not be produced for direct sale to consumers), but the taxi-like service the vehicle will provide.

"Our aim is to bring about the end goal of self-driving cars: this idea that you can summon a car to your doorstep, go safely in that car from A-to-B, get out, and pay a very low amount, if not free" he said.

The latter point ties in to remarks Cameron has made elsewhere suggesting that Voyage rides could be free and ad supported, although he declined to give any more information to Motherboard about the company's revenue model at this stage.

Where vehicle technology is concerned, the dramatic rise of Uber and Tesla (both of which became household names in just a matter of years) showed that disruption was still possible in the largely change-resistant auto market. But for a new company that aims to break into the taxi market in the US, competition from the Lyft/Uber duo will be stiff, and success will depend on Voyage's ability to distinguish itself and win over customers from the two.

Voyage is also planning to trial its vehicles on the road within the next few months, a task that brings regulatory complications of its own. So while there's lots of promise in the project, the cab firm still won't be in for an easy ride.