Intelligent Control Design for Quadrotor Perching Application using Neural-Network Augmented Direct Inversion Control Approach

Abstract-The quadrotor’s altitude(height), attitude(roll, pitch, yaw), and position (x-y directions) controller design are challenging research areas because of their non-linear coupled dynamics and under-actuated system architecture. This study proposes a quadrotor control system based on neural networks of the Elman recurrent learning mechanism. To solve the desired trajectory tracking problem for a quadrotor, a direct inverse control strategy utilizing Elman recurrent neural networks (ERNN) is demonstrated and tested through MATLAB simulation. The simulation findings show that the ERNN-based control systen operates with a minimum mean square error when using the reference flight testing dataset. Theerror-based comparativ analysis shows that ERNN-based altitude, attitude, and position controllers outperform the backpropagation neural network, according to our experiments.