Unmanned car map fighting is in full swing

Google is the undisputed winner in consumer digital maps and nowadays, driverless cars also need high-precision maps of the detailed data on the streets, and no one in the industry wants Google to see the last laugh. This map war battle is in full swing, many companies are competing.

No matter what day, there may be more than a dozen driverless cars in Silicon Valley drawing a map of the same corner. These cars come from different companies, but do the same thing: make high-precision street maps. These maps may eventually become a car navigation guide for unmanned vehicles.

These companies gather in places where law and weather are good – or where they can get the most attention. For example, a group of mapping vehicles gather near Las Vegas CES every year as the show is a popular stage to showcase the technological achievements of autonomous vehicles. Chris McNally, an analyst with Evercore ISI, said: “There could be 50 companies that ran a mapping of Las Vegas just to provide people with driverless car service during CES This is really a waste of resources. ”

Autonomous vehicles require powerful sensors to inspect the surroundings and also require advanced software to think and make decisions. In particular, they require the latest map of every imaginable road to travel. Who owns the most detailed and extensive map for vehicle reading, who will own a multibillion-dollar asset.

Because of this, a map war is opening in full swing: Dozens of competitors have entered a dazzling array of alliances that cost tens of millions of dollars in search of huge returns that may not come until many years later. Earlier in the year, Alphabet, Google, became the undisputed winner in consumer digital maps, where human pilots could bypass rush hour traffic or find restaurants. Google is how to win it? It dispatches cartography streetscape cars around the world, and it also has software expertise unmatched by map-based navigation companies, carmakers and even Apple. No one wants Google to win again.

So, companies involved in autonomous vehicle mapping are taking two different strategies, one of which is creating a complete high-precision map that will allow future autonomous vehicles to drive completely independently; the other is to create a bit by bit Map, the use of sensors on the vehicle, let the car gradually increase the degree of automatic driving.

Alphabet is taking both strategies. According to four insiders, a team within Google is developing a 3D map project that may license carmakers. This mapping service differs from the high-precision map that Waymo, another subsidiary of Alphabet, created for its autonomous vehicles.

Google’s mapping project focuses on so-called driver assistance systems to enable cars to automate some of their driving functions and help them see what is going on in the road ahead or at the corner. An early version of Google’s “Vehicle Map Service” released by Google in March incorporated sensor data into their maps.

Currently, Google is offering the service to car makers using Android Automotive, Google’s embedded operating system. So far, Google has created three partners for the system, but other car makers are reluctant to hand over their center console to the search giant. So, according to people familiar with the matter, Google is looking to expand the capabilities of the map service and look for other distributions.

In a statement, a Google spokesman said: “We’ve built a comprehensive world map for people, and we are working hard to expand the usefulness of our car-facing maps.” She declined to comment on plans for the future.

In the meantime, Waymo and other giants, including GM, Uber and Ford Motor Company, which have large unmanned research teams, each sent their own fleets to create rich and detailed high-precision map. There are also smaller startups selling devices or proprietary software that make such a map to car makers who feel they are far behind others. Still other vendors are building map services for general vehicles, but these services have only a handful of automation features such as adaptive cruise control or night vision.

These autonomous driving car maps are much more demanding than the old digital map requirements, resulting in huge investments in Detroit, Silicon Valley and China. “Autonomous cars want maps to be as accurate, accurate and real-time as possible,” said Bryan Salesky, who led Argo AI, a startup founded a year ago. “Off-the-shelf solutions do not exist.” Argo AI has won Ford’s $ 1 billion investment.

Cartographic tools trader

Just like creating a driverless car, making a driverless car map can be a daunting task. Unmanned test fleets, loaded with expensive lidar sensors and cameras, are to be dispatched with human drivers to capture the environment. Drawing the results helps to train the next team, the team will still be equipped with a safe driver – in some cases, you still need dozens of people sitting in front of the computer to classify all the camera.

This is a costly test and can take years to get a return, if not decades later. “How are you updating even though you can drive your own vehicle and travel on every road in the world?” Asked Dan Galves, a Mobileye spokesman. “You have to do it again Send these mapping vehicles. ”

Unlike traditional digital maps, autonomous vehicle maps require near-continuous updates. Even subtle changes in the road – overnight areas of buildings, or a bit of debris – may affect the driverless car. “Driverless cars do not work properly in the weird conditions,” said analyst McNally.

Mobileye believes it will be more efficient and cost-effective for the cars we drive today to see the roads ahead. In January, Intel announced a “low-bandwidth” mapping project that plans to bring its front-facing camera and chip sensor to 2 million cars this year. The company’s idea is to allow the car to see information such as lane markings, traffic signals and road boundaries so that they can be partially autonomous.

Mobileye said demand for computing power is less than building a comprehensive high-precision road map; Mobileye’s Greif said the company will use its sensor data to map with the navigation company to create a fully autonomous car Map that can be used.

This is also Google’s long-standing strategy of two map competitors, HERE and TomTom NV. The two European companies have positioned themselves as major alternatives to Google Maps and are dedicated to selling console maps to carmakers. However, these “static” maps can only see the rough street conditions and capture real-time snapshots. Now both companies are developing alternatives: “dynamic” maps that show lanes, curbs and everything else on the road. They hope that car makers stick to the products of traditional map makers as they transition from an intelligent vehicle to fully autonomous vehicles without a steering wheel.

HERE, owned by a German carmaker consortium, provides some examples. Its mapping system brings limited hands-free driving to Audi, one of its co-owners, and security capabilities this year for another co-owner, BMW, which acquired 15% of HERE last year.

Tesla, the company that supplies Autopilot, ancillary driving software, is the most urgent car company to embrace the strategy of gradually moving toward autonomous driving. Tesla relies on cameras and sensors on the vehicle but does not use lidar. The company has not disclosed which map service its Autopilot will adopt and its company representative declined to comment. Tesla unhappy with Mobileye two years ago parted ways to stop cooperation.

However, Tesla has at least chosen another company, Mapbox, to help integrate the map. According to the regulatory documents it submitted, Tesla paid $ 5 million in biennial licensing fees to Mapbox in December 2015. Mapbox mainly sells its location data to apps such as Pinterest and Snapchat. After a recent round of financing of $ 164 million, the company began to get involved in car maps. Mapbox says it’s capable of drawing about 220 million miles of road data globally every day, providing the latest snapshots of basic things like streets and lanes with software installed on the phone.

Eric Gundersen, the startup’s chief executive, said: “We currently have more sensors on the road than the total number of sensors in the connected automotive industry by 2020.” For automakers, , Mapbox’s selling point is to use its location data as a base layer for future maps – pairing it with camera systems like Mobileye’s or their own sensor data. Like other companies targeting carmakers, Mapbox is happy to play a neutral role and is willing to work with anyone. “We do not know who will win,” Gunderson said.

A new generation of Pathfinder

Not only is it unknown who will have the last laugh, the mapping industry does not even know which strategy is better and every autopilot car map will look different because different sensor systems will produce different maps . Nabeel Hyatt, an early investor in Cruise Automation from Spark Capital, said there is no standard sensor kit in the area. Cruise Automation was a self-driving car company bought by General Motors in 2016 for $ 580 million.

Therefore, a group of high-precision mapping companies are making different attempts on this issue. They are favored by venture capital and are also striving for lucrative contracts. Some of them disagree with Mobileye’s strategy, which relies on a seamless transition from semi-autonomous driving (so-called level 2 and level 3) to driverless driving (level 4 or 5). “It is very difficult to climb from Level 2 to Level 3 and then to Level 4 again,” said Wei Luo, chief operating officer of DeepMap Corp. “The gap is huge,” he said, with the best high Precision maps are built around driverless features. The startup said it is working with Ford, Honda and China’s SAIC, Mobileeye and SAIC, and Waymo is in talks with Honda about the deal.

Waymo is also in this camp. Originally known as Google’s driverless car project, which started as a map in 2009, Andrew Chatham of Waymo and another engineer started the “super tedious” drawing effort from scratch – Send a sensor-equipped car to capture the city’s surroundings and transform those 3D images into digital landscapes. Chatham said that the car may rely on the perception system to be able to drive on the highway, but not in other traffic conditions. Imagine a situation when you meet a busy double left lane intersection you’ve never seen before. Now imagine what happens to autonomous vehicles.

“This is the advantage of having a detailed map,” Chatham said. “We could give the car all the answers to the tough questions.” He said Waymo is exploring solutions that reflect real-time factors such as construction updates, but declined to give details.

With years of hard work and a strong library of artificial intelligence technologies, Waymo is considered a leader in high-precision mapping. But so far, the company has sold its entire suite to potential partners and achieved some cooperation. Chatham declined to say whether Waymo is considering marketing its map as a standalone product.

Another potential force in this market is Uber. The taxi service giant is also developing high-precision maps of its driverless car project, using a test vehicle similar to Waymo. Lisa Weitekamp, ​​Uber manager, said the company is exploring how to put map generation sensors in millions of cars on its service platform. Vittor Camp added that those maps already in use by the car – Uber’s “static” navigation software that includes popular routes and driving decisions – helped Uber create autonomous maps. “It gives us an edge,” she said.

That would make taxi service map access a valuable asset. Currently, Uber’s maps for its drivers and passengers use both TomTom, Google and its own data. Two insiders disclosed that the contract between Uber and Google will expire this year. Representatives of both companies declined to comment.

Many newcomers are trying to emphasize to car makers the need for automakers to catch up with leaders such as Waymo and Uber. DeepMap, co-founded by former senior employees of Google and Apple, relies on its smart software to reduce the time and cost required to convert images from autonomous vehicle sensors into a single high-resolution landscape. The startup said it is working with Ford, Honda and China’s SAIC.

Civil Maps has a technology that discerns sensor data that can be used to create a digital grid for each loop made by mapping vehicles in the same area. The company’s chief executive, Sravan Puttagunta, said it was a bit like the way mobile apps Shazam recognized a piece of music. Ford is an investor in Civil Maps and Putta Gangata says his company is raising additional funds.

At present, most car companies are at the trial stage and do not pay millions for the map. Ford spokesman described his cooperation with startups as “research.” Argo, a self-driving car the company is betting on, has learned about many suppliers but currently relies on its own in-house map. GM spokesman Ray Wert said the company prefers to make its own maps.

The new company is well aware that ultimately no one will survive. “This is very similar to navigable maps and even to search engines,” said Luo Wei, a staff employee who worked for Google at DeepMap. “Whoever has a bigger footprint will have an edge.”