Obstacle Avoidance with Energy Efficiency and Distance Deviation Using KNN Algorithm for Quadcopter

A Quadcopter is one type of UAV specifically widely used for search, rescue, reconnaissance, and others. The obstacle avoidance system in a navigation of quadcopter is needed to minimize human supervision. Obstacle avoidance systems can be created with dimension information of obstacle to choose avoidance directions. Due to power limitations in quadcopter flights, the choice of avoidance direction must consider energy efficiency. In this paper, the obstacle avoidance system in quadcopter navigation that flies in the 3D environment not only considers the dimensions of the obstacle, but also consumption energy and the distance between the quadcopter and the obstacle to choose avoidance direction (right, left or top). Efficient avoidance decisions are generated from KNN (K-Nearest Neighbor) machine learning with 96.6 % accuracy and require 0.0068s computing time. Simulation show that the quadcopter can reach the target point without colliding with static obstacles. Quadcopter can also choose the efficient avoidance direction when obstacles are detected.