Ford's unmanned car is in the same location in the lane, and every car is slightly deviated because they will avoid the hole. There is no problem with the car, the map is the root of the problem. The new map is perfect in the eyes Sensor of the human eye. But in the eyes of unmanned vehicles is not the case. A pixel has an error. This will cause the car to go wrong. The map shows its importance in new areas.

  In the past 10 years, when we walk or drive, the digital map for us to guide the basic direction. But technology and car companies have invested heavily in the development of a new generation of maps, and used in unmanned vehicles. These maps are often referred to as 'HD maps', which are far beyond the scope of navigation. Some companies continue to upgrade the map data, increase the lane mark line, road signs, traffic signals, hole information, and even including the height of the curb, the accuracy Fuel Rail Pressure Sensor of centimeters level. With such a map, the car can enter a more accurate world, the car will be more powerful sensor.
  When the lane mark line was covered by snow, or the truck blocked the car's vision, can not see the traffic signal, this time the map can guide the car driving. The map can also free the car sensors out, allowing them to focus on detecting objects that are not included in the map, such as pedestrians. In the road automatically driving the car, the map may not be so noticeable of the video, there is no laptops so eye-catching, Uber and Google confrontation court ABS Sensor is because the lidar sensor. But the map is still a key part of the puzzle. With a better map, it will be safer to deploy unmanned vehicles in the next few years, and it is possible to create a huge new industry.
 If you have an autopilot, the map is essential and not an optional feature. It will become the core component of the car, will create a continuous revenue stream. TomTom and other enterprises to sell traditional navigation systems, high-definition maps and navigation systems, for certain areas, high-definition map every day to update. For example, the frequency of updating data across busy urban streets and pedestrian areas across the construction area is higher than long, uninterrupted highways. Many companies are developing the next generation of maps, more and more intense competition. Google, Uber, Ford and other companies are research Throttle Position Sensor and development, want to use unmanned vehicle sensors to collect high-definition map data. Waymo is Google's unmanned subsidiary, confirming some of the targets, such as lanes and fire hydrants. Automatically send the report to the map team as the car changes. Even a strong business is not easy.
  In the first round of the map war, Google may be the winner, but the HD map has just developed. Who wins who lost no conclusion. Different from the ordinary electronic map used for navigation in daily life, there are two main characteristics of high-precision map for autopilot. One is to contain richer and detailed data information. These data are divided according to both moving and static. Static data includes both basic two-dimensional road data - such as lane markings, peripheral infrastructure, etc., and also covers quasi-static data such as traffic control, road construction, wide area meteorology. High-precision map also includes the accident, congestion and surrounding vehicles, pedestrians and lights and other rapidly changing dynamic information data. Unlike regular maps that are updated for several months or even years, high-precision maps must be kept minute-level and even second-rate updates.
 Second, positioning accuracy is higher. GPS navigation on the use of mobile phones, the accuracy Pressure Switch is generally in the 5-10 meters range, in the building area or underground tunnel accuracy even lower. While the high-precision maps required for autopilot technology will achieve centimeter-level accuracy. If you believe that no one will eventually appear in any place, then every city, every street to map, the details must be accurate. Look at the Uber unmanned car team, it only draws a map of the city, that is, Toronto.
  More and more intense competition, the industry began to think about a problem: these maps costly, on the road really practical? Some people think that cars will eventually become smart enough to implant depth learning techniques without having to rely on a lot of maps. The development of the algorithm, so that the car does not need high-quality map, Ford is exploring this possibility, but he completely abandoned the view of the views of doubt. When you drive to a place for the first time, the data is still saved, and then the second, third, tenth use, make it better.