The car is manufactured by the Phenikaa Group’s team, which runs on electricity, reaches 100 km a day with an average speed of 20 km / h, the battery fully recharges after 7 hours.
One morning at the school of Dr. Le Anh Son, Phenikaa University (Hanoi) started by having a self-propelled car that he and his colleagues built to carry out tests on sensitivity and accuracy. Step into the car, instead of starting with the key and moving with the steering wheel like a normal car, Mr. Son just selects the starting point and the destination first, press the button “self-propelled start”, the 6-seater car slowly rolling.
As the lead researcher, Dr. Son said, the car is equipped with level 4 autonomous features according to the Association of Automotive Engineers (SAE), such as 3D maps, Lidar sensors and high-resolution GPS, machine learning technology, deep learning. These features support four groups of systems including control systems, safety systems, identification systems, and intelligent control systems.
“There are currently two countries researching autonomous vehicles, Japan and the US, with the highest level 3 and 4 being tested. Vietnam builds the first self-propelled vehicle among a few countries that demonstrate the ability to master autonomous technology. onions, “said Mr. Son.
With his colleagues on the idea since July last year, Dr. Son used to think that this product was 90% failed because at that time there was no research on autonomous technology, as well as infrastructure (roads, smart lights). Every stage starts from scratch, “only the passion for technology is always ready”, Mr. Son said and determined that, if it fails, at least the team of teachers and students in the school will. can access new technologies or become open documents to support future research.
It took the first three months for him to find teammates, connect overseas Vietnamese scientists who had the same desire to develop autonomous technology in Vietnam. “Only after a phone call and some exchange information, the brothers doing research in Japan, Russia are ready to fly home to plan together,” said Dr. Son.
In the first stage, the team created a 3D map system so that the car could determine the starting position and the destination. The world has a number of open sources for reference, but only supports identification in small distances. The team had to find ways to change the algorithms themselves to suit the terrain and traffic characteristics of the country. Depending on the terrain, the system can scan into a 3D map of a 10-hectare area in just 30 minutes.
Two key elements of autonomous vehicles are autonomy and accuracy, aided by a system of sensors and recognition. The team uses lidar technology, laser-emitting sensors to help the car determine the current position, surrounding points and measure the angle of the steering wheel. In order to increase the accuracy of the vehicle, Dr. Son used additional GPS devices, to locate and report the reverse of the system, so that lidar recalibrated the vehicle’s position on the map with an accuracy of less than 1 cm.
Thanks to artificial intelligence (AI) technology, autonomous vehicles can know in front of a person or object.In the process of moving, the car automatically recognizes lanes, scenes, pedestrians and processes quickly. situations such as stopping at signs, pedestrians, curves, thanks to cameras and laser sensors. The more data you collect, the higher your vehicle’s visibility. Characteristics of people, terrain, and signs are processed on self-developed group AI model.
Dr. Son shared that as a vehicle in traffic, AI needs to identify 100% accurately, it is difficult to fully apply a model from abroad. Therefore, the group focused on building algorithms suitable for the actual conditions of the terrain, domestic traffic signs and the behavior of the driver.
He said, when the model was completed and pushed data to the central system, the 4G network speed was not strong enough, making it difficult for the group to transfer data. “This affects the accuracy of vehicle identification”, Mr. Son said. Finding an alternative solution, the team built an internal wifi system so that the signal was transmitted so quickly that the central unit could recognize and process everything going on in real time.
After 6 months of manufacturing and testing through tests, the group found that the car is sensitive to the object ahead, the braking immediately makes the occupants feel scared. In order for the car to handle smarter, he and his partner calculated the relative speed between two objects to determine the stopping distance and applied braking force of about 10%, the vehicle brakes softer.
According to Dr. Son, the product has been completed and continued to be tested and improved to reduce costs. The team plans to shorten some of the sensors in the car but still ensure accuracy and safety by improving the algorithm strongly enough. As a result, the car has the ability to decelerate itself when cornering, obstacles ahead. With an average speed of 20 km / h. In the case of exceeding the permissible speed, the system will signal an alarm, the user presses the emergency stop button.
“Currently, there are no regulations or documents under the law on autonomous vehicles in traffic, so the group wants to bring products to the resorts and golf courses,” he said. In addition, the group continues to research to apply autonomous technology, create many new smart products and devices.
Self-driving cars are divided into 5 main levels:
– Level 1 : the simplest level, the driver must master the steering wheel, the vehicle has additional minimum assist technology, such as cruise control, lane keeping assist, parking assist …
– Level degree 2 : at this time, the driver no longer has to worry about the gas, brake, steer. Technologies on the car such as automatic braking, lane change assistance, adaptive cruise control … However, the driver still has to focus on the journey, ready to control. This is a popular level on cars today.
– Level 3 : the car is more active, no longer stops at “support” but can be active, such as automatically changing lanes …
– Level 4 : almost completely self-driving, the driver does not need intervene, but still have to pay attention.
– Level 5: the driver does not need to pay attention to the vehicle.
Follow vnexpress