The technology used in aircraft, especially supersonic fighter jets, is improving day by day, and their efficiency and speed are increasing significaly in each generation. But the question is, can the pilots also adapt to these developmes? Excessive speed puts a lot of pressure on the pilots’ body and can lead to loss of consciousness and even death. Cases like this cause the lack of human capacity to use new technologies.
“Autopilot” is a solution that is considered as a replaceme for human pilots. These iellige autopilots have operated as co-pilots in certain cases; But due to many defects, they have never managed to fully corol an aircraft. A challenge that MIT researchers are trying to overcome by preseing a new solution based on machine learning and take an impressive step towards achieving fully autonomous aircraft.
The researchers’ new approach has been able to increase the safety and stability of automatic pilots up to 10 times and guide the plane to the target poi in complete health. Chuchu Fan, an assista professor of aeronautics and one of the researchers of this project, says: “Previously, many people were ierested in working on the problem of fully iellige pilot; But when faced with such a large and complex aerodynamics, they didn’t know how to deal with it.”
The machine learning model used in this research is based on the trial and error method. In this model, every correct behavior of the machine to achieve its goal is met with a reward, and after thousands of attempts, the desired modeling is done. But the challenge facing the researchers in this technique is that two goals are considered for the automatic pilot; Stay stable and avoid obstacles. To solve this challenge, they divided the experime io two differe parts.
In the first part, the optimization problem is limited to staying on a certain path. So that the autopilot is trained to stay on a predetermined path and take the plane to the iended target. In the second stage, obstacle avoidance is implemeed and the aircraft must perform extreme maneuvers to avoid hitting obstacles in the middle of the road.
Dr. Su, another researcher of this project, says: “Our goal is for the iellige pilot model to reach a level of safety and stability that can take over emergency manageme in difficult situations when even a human pilot is unable to solve the problem. And come out of it proud.”
The implemeation of this advanced model in the real world could be a turning poi among self-driving vehicles. Because even if they are not completely erusted with the guidance of passenger planes, they can be useful in corolling private cars and helping to maiain the balance of the car on snowy and slippery roads.
What do you think about this? Will there come a day when the corol of airplanes will be completely handed over to artificial ielligence?
