Shahrood University of Technology Journal of Solid and Fluid Mechanics 2251-9475 14 1 2024 03 20 Designing, Construction and Attitude Identification of a Portable Quadcopter Using Flight Test Data Based on Genetic Algorithm and Artificial Neural Networks Designing, Construction and Attitude Identification of a Portable Quadcopter Using Flight Test Data Based on Genetic Algorithm and Artificial Neural Networks 93 107 3171 10.22044/jsfm.2024.13843.3808 FA Meysam Delalat Aerospace Engineering, Faculty of Aerospace Engineering, Malek Ashtar University of Technology, Iran Alireza Ahangarani Farahani Malek-Ashtar University of Technology Seyed Majid Hosseini Malek-Ashtar University of Technology Journal Article 2023 11 18 The current research has two main parts, the first part is the process of design, manufacture and selection of quadrotor's elements and the second part is modeling and estimation of model parameters based on flight test data. The main approach of making this drone is to be light and cheap. Also, it is very useful to develop a precise mathematical model for the quadrotor to control the system. Therefore, a system identification method for the quadrotor model based on genetic algorithm and neural network is investigated in the second part. The advantage of system identification based on experimental data is designing a better controller. Using flight test data including angles and rate of them, the identification of the system based on the two methods of genetic algorithm and neural network has been done. For genetic algorithm modeling, a linear second order function is considered for each channel. Meanwhile, it is assumed that the channels are decoupled from each other. The results show that in the roll axis, the results of the identification are the same, in the pitch axis, the accuracy of identification with the second order function and the genetic algorithm method has better results, but in the yaw axis, the accuracy of identification with the neural network is significantly better. In general, the adaptability in all three axes is 81.48% for the genetic algorithm method and 82.62% for the neural network algorithm method. The current research has two main parts, the first part is the process of design, manufacture and selection of quadrotor's elements and the second part is modeling and estimation of model parameters based on flight test data. The main approach of making this drone is to be light and cheap. Also, it is very useful to develop a precise mathematical model for the quadrotor to control the system. Therefore, a system identification method for the quadrotor model based on genetic algorithm and neural network is investigated in the second part. The advantage of system identification based on experimental data is designing a better controller. Using flight test data including angles and rate of them, the identification of the system based on the two methods of genetic algorithm and neural network has been done. For genetic algorithm modeling, a linear second order function is considered for each channel. Meanwhile, it is assumed that the channels are decoupled from each other. The results show that in the roll axis, the results of the identification are the same, in the pitch axis, the accuracy of identification with the second order function and the genetic algorithm method has better results, but in the yaw axis, the accuracy of identification with the neural network is significantly better. In general, the adaptability in all three axes is 81.48% for the genetic algorithm method and 82.62% for the neural network algorithm method. https://jsfm.shahroodut.ac.ir/article_3171_f41967fa7774d67b291d0477c231e81e.pdf