MIT’s new algorithm can train drones to relax and avoid obstacles to fly faster

Drone competitions are a relatively new sport that typically compete on tracks with obstacles, and drones need to avoid obstacles as quickly as possible. Although drone competitions are purely for fun, the technology of avoiding obstacles in these competitions can also allow commercial drones to avoid obstacles during complex situations and time sensitive operations, such as search and rescue. The Massachusetts Institute of Technology is attempting to enable drones to fly faster while avoiding obstacles.

MIT aerospace engineers have developed an algorithm that allows drones to choose the fastest route to bypass obstacles without colliding with them. This algorithm combines simulation of drones flying over virtual obstacles and experiments involving actual drones flying the same route in the real world. When drones are trained using new algorithms, their speed of crossing obstacles can be 20% faster than drones planning flight routes using traditional algorithms.

Although the MIT team found that their algorithm can significantly accelerate the speed of flight through courses, they also found that drones trained with their new algorithm are not always faster than those trained with traditional algorithms. In professional racing, both pilots and drivers know that sometimes you have to slow down in certain areas to go faster in other areas.

The algorithm of MIT can determine that in order to travel faster on the subsequent route, even if it is surpassed by competitors, it is better to slow down. The project researchers believe that the algorithm they have developed is an important step in enabling future drones to browse highly complex environments very quickly. For example, this technology could one day be used for unmanned aerial vehicles in search and rescue operations to quickly and accurately navigate crowded and dangerous environments.

Prev: Machine learning enables further and more accurate analysis of orthopedic images