MIT is out of black technology again! The unmanned vehicle is equipped with ground penetrating radar, which can identify roads with different materials.
Text || James
On March 3 rd, researchers at MIT recently proposed a solution to the failure of autopilot system in bad weather. They installed the commonly used "ground penetrating radar" on the self-driving car, and the car realized automatic driving by detecting the situation under the road. Even in bad weather, the error of actual road test is very small.
At present, many self-driving vehicles that have been put into test are equipped with lidar, sensors and cameras to detect the road ahead. However, in rainy, snowy and foggy weather, even the best lidar, sensor or camera can hardly play its full role. This is because the "eyes" of the sensor are also covered in rainy and foggy days, and it is impossible to obtain a clear road image.
Researchers at the Massachusetts Institute of Technology have solved the problem of the decline in the recognition ability of the autopilot system in bad weather with another way of thinking. However, WaveSense, which produces this detection system, is expected to start the actual road test this year, and there is no market launch plan for this product at present.
1. Can "ground penetrating radar" replace lidar in bad weather? The actual measurement requires almost no human intervention.
Recently, the research team of Computer Science and Artificial Intelligence Laboratory (CSAIL) of Massachusetts Institute of Technology said that if a new "GPR" system is installed for self-driving vehicles, the road detection ability of self-driving vehicles in bad weather will be improved.
As the name implies, "ground penetrating radar" is a radar system that can detect underground conditions. The system can send electromagnetic pulse to the lower part of the road when the vehicle is driving, and then receive the feedback signal from the lower part of the road. Because the feedback signals of rocks, mud and ordinary urban roads are different, the system can judge whether the vehicle is driving on the right road.
▲ Working principle of "ground penetrating radar"
Researchers at the Massachusetts Institute of Technology said that after studying different substances under the road, it was found that each substance can send out different feedback signals under the signal of electromagnetic pulse, so the system can identify different road conditions. Once the camera or lidar system fails in bad weather, this "ground penetrating radar" can help self-driving cars continue to detect road conditions.
Teddy, a doctoral student in the Computer Science and Artificial Intelligence Laboratory? Ort said: "The ground penetrating radar system can quantify various substances under the road surface and compare these data with the built map. In this way, the position of the vehicle on the road can be known without the assistance of a camera and a lidar. "
In addition, the accuracy of ground penetrating radar is very high. In the test, the researchers found that the error of "ground penetrating radar" in snowy weather is only 1 inch (about 2.54 cm) compared with sunny days. The error of rainy weather is slightly larger, but it is only 5.5 inches (about 13.97 cm).
The researchers said that they installed the "ground penetrating radar" system on self-driving cars for testing, and almost no human intervention was needed in the test. However, at present, MIT researchers have only conducted tests on rural roads with low speed limits, and have not conducted urban roads or expressway tests.
Second, "ground penetrating radar" has been used for all kinds of underground detection? But it will not be put on the market for the time being.
Professor Daniela from MIT? Ruth said: "Our work has proved that this method can make autonomous vehicles continue to work in bad weather, and you can' see' the road without using lidar or camera."
Researchers believe that in addition to allowing self-driving cars to "see" the ground cleaning situation, with the passage of time, the map detected by "ground penetrating radar" will have more advantages. For example, the map data generated by "ground penetrating radar" system is 20% less than that generated by traditional 2D sensors, which can save more storage space.
Previously, "ground penetrating radar" technology was often used for lunar exploration, mine detection and urban construction planning. In 20 18, the self-driving company WaveSense released the "ground penetrating radar" system for self-driving cars, and the "ground penetrating radar" officially entered the field of self-driving. WaveSense is expected to start testing this year.
▲ The working scene of "Ground Penetrating Radar" developed by ▲WaveSense Company.
However, there is no plan to put the ground penetrating radar system on the market at present. Because the commonly used radar sensors such as lidar and camera are already huge, if a "ground penetrating radar" is added, the volume of the sensor will be expanded and the aesthetics of the vehicle will be reduced. In addition, the road map and the underground map are completely different, and the cooperation between the two maps will also be a difficult point in the future.
Conclusion: GPR is a good solution? Still a burden?
At present, the automatic driving system is developing rapidly, but it is also difficult to identify the road in bad weather. This is also one of the "pain points" of autonomous driving that major car companies and technology companies focus on.
"Ground penetrating radar" is a good solution, but it can only detect road conditions, but it lacks the detection of pedestrians and vehicles on the road, so it needs more hardware equipment to cooperate. This increases the complexity of autonomous driving to a certain extent, and also puts forward higher requirements for the computing power of autonomous driving equipment. Whether the "ground penetrating radar" is really practical needs further study and exploration.
This article comes from car home, the author of the car manufacturer, and does not represent car home's position.