Designing a reinforcement learning-based robust controller resisting external forces for a octarotor with new structure

چکیده مقاله

Most of the researches conducted in the multirotors field can be divided in two parts of structure and layout of the new flying propellers and designing controllers fit for the desired performance. One of the issues stated in this study is the limitation of independent control of 6 DoF for the aerial vehicle with conventional structure such as octarotors. Conventional aerial vehicles are only capable of controlling in 4 DoF. This is because all rotors are placed in one plane and the resulting thrust force is parallel. The presented paper aims to create a novel structure in the configuration of the rotors of an octarotor to realize the capability of controlling the position and angle of the octarotor independently and going through a predetermined path and being in specific directions. The proposed model offers the capability of changing the octarotor speed and angle of the four arms for controlling purposes. Also, to optimally provide the maneuverability of the aerial vehicle a robust controller using neural networks and reinforcement learning is proposed to determine these velocities and angles. Therefore, the ability to control the octarotor with 6 DoF independently, ability to track path and statues of the octarotor, and also, robustness against external forces, variation in payload weight, and possible disturbances in motor performance are the results of this study. Finally, simulation results for the mentioned abilities are presented and examined.