Recently, Toshiba announced that in the field of LiDAR lidar for distance measurement, it has developed a technology that can track vehicles, people, and other objects with 99.9% accuracy, achieving the world's highest accuracy. And only using LiDAR to collect data can achieve 98.9% object recognition.
In addition, the detection distance in rainstorm and dense fog environments has been increased by more than twice, and a "rain fog removal algorithm" has been developed, which can measure the distance of 40 meters in a violent rain environment of 80 mm/h. In addition, through the "variable measurement range technology", compared to existing technologies, the measurement distance can be extended to 350 meters, while achieving a measurement distance of up to 120 meters, and the viewing angle has been increased by about 6 times. All three technologies are planned to be put into practical use in 2025.
In addition to being used for autonomous driving, LiDAR also plays an indispensable role in monitoring social infrastructure such as roads and railways. By improving the accuracy and flexibility of the measurement range, the company stated that it will contribute to the progress of autonomous driving and the efficiency of social infrastructure by building a digital twin.
searcher at the Internet of Things Edge Laboratory at the Information and Communication Platform Research Institute of Toshiba Research and Development Center, said: The challenge of LiDAR is that it is difficult to accurately identify objects and track them using only LiDAR measurement data. When used in conjunction with cameras to measure 2D and 3D data, there is a spatial deviation between the two types of data, leading to a decrease in accuracy. In addition, there are also issues such as decreased accuracy in harsh weather environments such as rain and fog, as well as blind spots caused by installation location limitations. Toshiba's new technology can solve these problems.
The first technology is "2D/3D fusion AI", which can accurately recognize and track objects using only data obtained from LiDAR. So far, it is necessary to match the perspective and frame rate of the camera and LiDAR and accurately synchronize them. However, in addition to correcting errors, the perspective and frame rate may also shift due to factors such as vibration, resulting in a significant decrease in recognition accuracy. 2D/3D fusion AI only uses LiDAR to obtain 2D and 3D data. By integrating the two and applying them to artificial intelligence, high-precision object recognition and tracking can be achieved through learning.
Cui said, "Toshiba's LiDAR has the characteristic of being able to obtain both three-dimensional data and two-dimensional image data. Since only one LiDAR is used to collect data, two-dimensional and three-dimensional data can be read from the same pixel at the same time without the need for merging, so there is no need to worry about the decrease in recognition accuracy." This way, for any vehicle and pedestrian within a distance of 50-115 meters from LiDAR and 80-110 meters from LiDAR, there is no need for cameras, Even at night without lighting, 98.9% of vehicles and people can be recognized and 99.9% can be tracked. Additionally, the tracking error is limited to 15 centimeters.
The second "rain and fog elimination algorithm" minimizes the decrease in LiDAR measurement accuracy caused by rain and fog. The light emitted by the infrared laser used in LiDAR has the property of absorbing and scattering when encountering moisture. In outdoor environments with poor visibility such as rain, fog, and snow, there are problems of reduced measurement accuracy and shorter detectable distances. In the newly developed technology, an AD converter is used to convert analog data into digital data. Based on the digital value of reflected light intensity, the characteristic quantity of reflected light generated by scattered particles such as water is used to measure whether it is rain, fog, or cars. If it is judged as rain or fog, deleting the waveform can extract fragile reflected light covered by rain or fog, as well as reflected light from the measured object.
In experimental equipment simulating real environments, the detectable distance was increased from 20 meters to 40 meters in a severe rainfall environment known as a disaster level of 80 millimeters per hour, and from 17 meters to 35 meters in a fog environment with visibility of 40 meters, both achieving more than twice the increase.
The third type is the "variable measurement range technology", which determines the distance and angle of view of LiDAR based on the set location, allowing for the freedom to change the measurement range. Toshiba has announced LiDAR technology, which uses two miniaturized floodlights to increase the measurement distance by 1.5 times. This technology further extends the measurement distance and increases the viewing angle by 6 times by changing the number of floodlights and the configuration of the receiving lens.
At a field of view angle of 60 degrees (horizontal) × Achieved a world-class measurement distance of 120 meters with a field of view angle of 24 degrees (horizontal) at 34 degrees (vertical) × At 12 degrees (vertical), the world's longest measurement distance of 350 meters was achieved. In addition to monitoring infrastructure such as roads and tracks that require remote measurement, it can also be used for automatic driving of AGVs (Automatic Guided Vehicles) in factories and warehouses that require wide-angle performance, thereby accelerating the realization of spatial digital twins.
Cui stated that Toshiba will further promote the research and development of environmental resistance performance, with the goal of achieving commercialization of solid-state LiDAR by 2025. We envision the widespread application of LiDAR in mobile automation, high-precision infrastructure monitoring, and the construction of spatial digital twins.
Cui also introduced the use of LiDAR for spatial digital twins. Digital twin is an analog technology that utilizes sensing technology and artificial intelligence to link real-time changes in the real world, allowing for real-time collection of data such as operating devices, faithfully reproducing events that occur in virtual space. In recent years, digital twins have not only targeted specific devices, but also spaces such as factories, warehouses, stores, and roads. The spatial digital twin can accurately capture all the movements and stillness of people, objects, and things on site in real time, and will bring benefits in various applications. For example, even without people patrolling, the digital twin can detect situations such as road collapse, icing, and cargo falling. Based on this information, staff can quickly start removing work and make decisions, For example, prioritize operations in areas that may cause large-scale congestion. In addition, when starting a crane at a construction site, even if it is impossible to confirm whether there are people in the blind spot of the actual construction site, the danger can be detected in real time and notified to the site through spatial digital twin equipment.
Cui metaphorically said, "LiDAR has a distance error of 5 centimeters beyond 50 meters, and due to its own luminescence, it can measure distance without illumination. Therefore, by constructing a spatial digital twin that correctly understands the condition of all things, identifying and judging the best state, human error can be eliminated. All places will be protected by the 'Eye of God'." He also pointed out that spatial digital twins will accelerate the popularization of automation, Including autonomous driving. For example, on highways, the implementation of spatial digital twins can support 18 trucks driving automatically at a distance of 10 meters and a speed of 60 kilometers per hour.
In addition, by using quantum computers to analyze the data that constitutes spatial digital twins, the ideal scenario will be reproduced in the metaverse. It is possible to grasp the differences and anomalies between the ideal form and eliminate dangerous safe and efficient driving. "With the development of LiDAR, in addition to the use of automatic driving, it is also possible to build spatial digital twins and promote the overall optimization of all mobile automation, manufacturing and logistics lines in space. By making great contributions to the construction of spatial digital twins using LiDAR, it is possible to realize the automation of industries shaken by COVID-19 disasters and lack of manpower."
Source: Yangtze River Delta Laser Alliance
