Are you familiar with Quora.com? That’s a website where people can ask a question and get an answer from a knowledgeable person. Recently, this question was posted there: “Despite the efforts made by Google to get its self-driving cars on roads, Tesla launched it. Is Tesla better than Google in this field? ”
That’s a really good question. Google has been developing self driving cars for years. It has a fleet of Lexus RX450h cars equipped with cameras and scanners crisscrossing Silicon Valley, gathering data that will allow them to drive themselves without input from a human driver. (The state of California requires that there be a real person on board, ready to take over control of the car at a moment’s notice.)
It also is working hard on the Google Car, a cute little bubble of a vehicle that may offer insight into the future of mobility. Instead of millions of privately owned cars fighting their way into and out of cities every day, a smart phone app could summon a self driving car to your location at any time and take you wherever you need to go for a modest service fee.
No car payments, no taxes, no registration hassles, no repairs or maintenance, and no insurance bills. That actually sounds pretty appealing, doesn’t it?
The question on Quora elicited a detailed and insightful answer from Mike Barnard, a senior fellow at the Wind & Energy Institute. He is also a contributor at our sister site, CleanTechnica. Are you ready to hear more about Mike’s answer? Find yourself a comfortable chair, put your feet up, and relax. This may take a while.
As reported by Forbes, Mike begins by saying he has a background in robotics. He says there are two basic ways a robot can find its way from Point A to Point B (and maybe Points C, D &E).
- The first is a full world map, in which the robot or a connected system has a complete and detailed map of the world and a route is planned along that in advance accounting for obstacles. Basically, the robot has to think its way past or over every obstacle, which makes for a lot of programming.
- The second is subsumption architecture robotics, in which a robot is first made so that it can survive environments it will find itself in, then equipped with mechanisms to seek goals. The robot then, without any idea of the map of the world, navigates toward Point B. The robot is robust and can stumble its way through obstacles without any thinking at all. The original Roomba vacuum cleaner was a pure subsumption beast.
Google uses the full world map model. Barnard says, “For one of its cars to work, [Google] needs an up-to-date, centimeter scale, 3D model of the entirety of the route it will take. Google’s cars are ridiculously non-robust and when confronted with something unusual will stop completely. Basically, all intelligence has to be provided by people in the lab writing better software.”
The full world map model requires prodigious amounts of data to work. Acquiring that data can take years, if not decades. Why would Google down that road, so to speak? Barnard says it may be the result of Google’s corporate culture. It’s entire business model is predicated on amassing reams of data and building sophisticated algorithms to manage and interpret it. In other words, instead of thinking outside the box when it comes to building autonomous cars, it built its own box then locked itself inside.
Tesla, on the other hand, begins with the subsumption model and then improves on it, using what Barnard calls “intelligent real world research assistants.” They are the drivers of Tesla automobiles who add “focused, experienced instincts into its cars.” Every action those drivers take is uploaded to the Tesla cloud. From there, it is shared with every other Tesla connected to the cloud. If you travel a certain road that has a posted speed limit of 55 mph, that information becomes part of the knowledge basis contained onboard every other Tesla.
If a speed limit sign goes missing, every Tesla travelling the same bit of road will know what the speed limit is. If one Tesla figures out how to navigate despite an ambiguous exit sign or on a road with no lane markings, every other Tesla will know how to do that as well. A Google Car can’t do those things. It has to wait for someone to update its software.
Barnard points out that within days of Tesla releasing its autonomous driving software — known to Tesla aficionados as firmware version 7.0 — a group of intrepid Tesla pilots crossed the entire continent in less than 60 hours, with the Tesla driving itself 90% of the time. For a Google Car to do something similar, it would first have to send out a mapping car to plot the route. Then that data would have to be shared with the self driving car, which would follow along a week or so later.
The beauty of the subsumption model used by Tesla is that its cars do not require a world map that is exact to the centimeter. Barnard says that one factor alone means that Teslas are “just fine with much coarser grained maps which are much easier to build, store, manipulate and layer with intelligence as needed.” He concludes with this thought: “The rapid leaps in capability of the Autopilot in just a few days after release should be giving Google serious pause. By the time its software geniuses get the Google car ready for prime time on a subset of roads that it has mapped to centimeter scale with special sensor cars, Teslas will be able to literally drive circles around.”
And that is why Tesla will rule the world of autonomous driving. Thank you, Mike Barnard, for your expert analysis.