Development of a deep reinforcement learning model for UAV motion planning during windy conditions.
Abstract
Over the past years, UAVs have garnered high application in the civilian space as
they have been used in package delivery, wildlife monitoring, disaster management,
among other applications. This increase in use has in turn led to increase in quantity
of drones in the air space.
In this research project, we utilized deep reinforcement learning to enable a UAV
carry-out autonomous
ight in an area with static objects during windy conditions.
In the simulation environment, the UAV will be tasked to move to a target location
in an environment with variable wind speed and has static objects. The drone is
equipped with a front camera that can continually take 640x480 pixels that will
be enable the UAV identify the objects in the environment hence taking favorable
actions to avoid collisions.
In addition to the UAV image, drone battery, heading angle of UAV and distance
of the destination from the position of UAV was added to the state. Therefore, we
used a concatenated neural network for the deep reinforcement learning algorithm.
Finally, as you will read later, our key fi nding was that the reinforcement learning
model we developed bettered the other path planning algorithms in terms of battery
percentage left.