Unmanned vehicle Sensor GPS In-depth analysis _ unmanned driving

Turn from: https://zhuanlan.zhihu.com/p/32638633

In the last share, I made an introduction to the technical framework of Baidu Apollo plan, as shown in the picture. Apollo1.0 automatic driving on closed site

If you want to complete the Apollo 1.0 "closed site tracking Automatic driving" function, we need to solve an important problem: I (unmanned vehicle) where.

"Where I am" is completely dependent on the GPS (Global Positioning System) and IMU (inertial measurement unit) in the Apollo 1.0 architecture.

Today's sharing, I will use the language as simple as possible to introduce the principles and characteristics of GPs, and talk about why the positioning of unmanned vehicles not only rely on GPs, but also rely on IMU. I will focus on the function of IMU in the next sharing. Body

The principle of GPS positioning is simple

GPS is the abbreviation of Global Position system, which is commonly used in automobiles and mobile phones. Civilian GPS positioning accuracy between 10~20 meters, this is why the mobile phone GPS positioning, sometimes clearly on the land, but is thought to be in the river reason.

The principle of GPS positioning is very simple, called triangulation method (triangulation). Principle diagram: Picture source: https://www.nationalgeographic.org/photo/triangulation-sized/

The GPS receiver installed on the unmanned vehicle, first measure the radio signal to reach the satellite transmission time, and then multiply the transmission time by the speed of light, you can get the current GPS receiver to reach the satellite distance, with a distance, it is possible to obtain the position according to the geometrical principle.

Photo Source: http://slideplayer.com/slide/5717261/

If the GPS receiver is known to reach the distance between the 1th and the number 3rd, the 1th and 3rd satellites produce two spheres (green and blue) from distance, and the intersection of two spheres is circular, and any point near the Earth's surface is likely to be the position of the current unmanned vehicle (GPS receiver).

It is therefore not possible to determine the exact location of the current unmanned vehicle based on these two distances alone. By introducing the distance of the third satellite, the location of the unmanned vehicle can be accurately determined. This is the principle of triangulation method.

GPS positioning accuracy is not enough

According to the above, as long as the radio signal reached three satellite transmission time, it seems to be able to calculate the location of the unmanned vehicles.

But engineering problems are never as simple as theoretical calculations.

In the open and thin clouds, the transmission time of the radio signal times the speed of light is indeed close to the actual distance.

But the transmission time of the radio signal will be affected by the media. such as the thinning of the clouds, weather and so on will affect the transmission time, and thus affect the distance calculation. Distance is not allowed, the result of the positioning of course is not allowed.

In order to reduce the weather, the impact of cloud on positioning, engineers introduced a differential GPS technology. As shown in the following figure: Photo Source: http://what-when-how.com/gps-with-high-rate-sensors/carrier-smoothed-code-gps/

Base station is built on the ground (base station), when the base station is established, the precise position information (latitude and longitude) of the base station can be obtained. At the same time, the base station has the function of receiving satellite signal.

When the GPS receiver of the base station is close to the vehicle GPS receiver (<30km), it can be concluded that the GPS signal of the two is in the same atmosphere region, that is, the signal error of the two is basically the same. According to the precise position of the base station and the time of the signal propagation, the signal propagation error caused by the weather is reversed, then the GPS signal of the vehicle is corrected by the error, which can reduce the influence of the cloud and weather on the signal transmission.

Using the differential GPS technology, the positioning accuracy of unmanned vehicle can be improved from 10 meters to the meter level.

Differential GPS can solve the problem of positioning accuracy, but it can't solve the problem of occlusion and reflection.

When the unmanned vehicle is traveling in the high-rise environment, this will have the following problem. Photo Source: https://www.e-education.psu.edu/geog862/node/1721

1. Signal loss

GPS receiver in high-rise buildings around, it is easy to lose a certain direction, all the satellite signals, rely on the other three of satellite signals to obtain the positioning results, in the precision is difficult to meet the needs of unmanned driving.

2. Multi-Path problem

Satellite signals that could not have been received in the vicinity of high-rise buildings are received by specular reflections from the building's body building, known as multipath signals (Multi-path Signal). It can be seen from the graph that the distance calculated by multipath signal is significantly larger than the actual distance. And the unmanned vehicle is very difficult to judge, the current receives the signal is the single path or the multipath.

In view of the above reasons, it can be seen that the GPS sensor alone, the unmanned vehicle in the complex scene, it is difficult to achieve precise positioning.

GPS positioning frequency is not too high

GPS positioning frequency is not high, only 10Hz, that is, 100ms to locate once.

With a simple example, tell me why 10Hz is not high enough.

Assuming that a car is in the 72km/h (20m/s) speed on the road, GPS positioning cycle of 100ms, after a period of time, the car relative to the previous location of the results of the movement of the 20m/s * 0.1s = 2m, that is, two times the location result is 2 meters apart.

If the unmanned vehicle is travelling on a path with curvature, it means that every 2m can be based on the location of the car, a control (steering wheel angle, throttle opening, etc.) calculation. The driving effect of the vehicle under this control frequency is shown below. The unmanned vehicle is left and right on both sides of the actual trajectory and cannot travel precisely along the track.

In contrast to the ideal control method, unmanned vehicle trajectory should be the following figure, maintain extremely high frequency of positioning and control, each take a small step, are again control calculation and execution control.

In order to solve the problem caused by the low GPS frequency, engineers introduced other sensor signals (IMU, laser, vision) to improve the positioning frequency of unmanned vehicles.

How these assistive devices can be realized will be shared with you later.