In autonomous vehicles, AI determines the surrounding situation and drives the vehicle based on information obtained from various sensors installed throughout the vehicle. When using the autonomous driving function, it uses a sensing technology called LiDAR (Light Detection And Ranging) that monitors all directions in 360 degrees in addition to the camera sensor, which is used to recognize lanes or signs.
The rider irradiates a laser pulse to detect surrounding objects based on the time it takes to receive the reflected light. However, in such a structure, for example, roads with snow piled up are buried or covered in thick fog, and sometimes the accuracy of recognizing roads and surrounding objects is significantly reduced due to lighting conditions.
MIT CSAIL has developed a method of combining technology called GPR (Ground Penetrating Radar) to address these rider weaknesses. Because GPR uses high-frequency electromagnetic pulses, cameras and riders can detect conditions on the ground, such as lighting conditions, snow, and soil and rocks, where precision is difficult to expect. The data can then be converted into a map for autonomous vehicles to identify roads.
The GPR system, called LGPR (Localizing Ground Penetrating Radar), developed by MIT and the Lincoln Research Institute established by the U.S. Department of Defense, detects the condition of the soil without worrying about snowfall or thick fog, and determines the road situation by matching with a prepared map. do. CSAIL tested the system at low speeds by shutting down rural roads with snow accumulation, but basically expects to be able to easily expand the application of the system on all roads, including highways.
However, it is not without weaknesses. LGPR is not well aware of the state of water seeping into the floor due to rain. Since objects on the road cannot be measured, LGPR alone cannot be used for autonomous driving and must be combined with other sensing devices. Research papers on autonomous driving using LGPR are published in the journal (IEEE Robotics and Automation Letters). The research team plans to make the current 1.82m wide equipment compact and continue improving the LGPR mapping technique. Related information can be found here .