Machine Learning Algorithm will train drones to overcome obstacles

Drones were invented to assist soldiers during wartimes. Though drones still serve the primary purpose in wars they are now used in a variety of realms. They are used to provide aid during natural disasters, for helping medical professionals in drug delivery, for remote sensing, and oil mineral exploration. 

 

A drone’s maximum speed can be 100 mph. Autonomous drones that operate via artificial intelligence systems  search out for the fastest routes to help people in dire need. Drones also map their own routes to overcome obstacles in the way. 

 

Deciding the right route can take a lot of time and it also affects the functioning of the drone. Halt in the drone operations can have serious consequences on some people’s lives. 

 

To overcome these problems , aerospace engineers at MIT developed a machine learning algorithm that will help autonomous drones to hunt out the fastest route to overcome the obstacles that might come in its way. 

 

Newly devised algorithm is a result of two combination studies. First study takes into consideration data from the virtual obstacle course. Second study takes into account data related to drones' physical obstacle encounters. 

 

Autonomous Drones trained with new machine learning algorithm

 

Researchers concluded the study, with the conclusion that drones trained using the machine learning algorithm were 20 percent faster in locating a shorter route. Drones when encountered by an obstacle followed the slowing down or changing the direction strategies. 

 

(Must Check: 7 Applications of IoT in Defence and Military)

 

To put the model on display researchers put forward five large square shaped obstacles in the way of the drone. These obstacles were placed in the physical training space. Drone was programmed to follow a specific speed. Drones trained on new algorithms were placed in a competition with conventionally trained drones. 

 

Machine learning algorithms trained the drone to take turns to avoid collision with the obstacle. It followed the same speed from the beginning but performed better as it was trained to overcome obstacles. 

 

Drones will be trained in more tricky environments for the future. Real data will be evaluated for better drone training. 

 

According to MIT News Sertac Karaman, associate professor of aeronautics at MIT said, “These kinds of algorithms are a very valuable step toward enabling future drones that can navigate complex environments very fast.” “We are really hoping to push the limits in a way that they can travel as fast as their physical limits will allow.”