Intelligent control of an UAV with a cable-suspended load using a neural network estimator

• Published in: Expert systems with applications 2021-11, Vol.183, p.115380, Article 115380
• Main Authors: Sierra-García, Jesús Enrique, Santos, Matilde
• Format: Article
• Language: English
• Subjects: Artificial neural networks; Cable-suspended load; Complex systems; Complexity; Feedback linearization; Hybrid automata; Hybrid control; Hybrid systems; Intelligent control; Load; Modelling; Neural networks; Robust control; Stability; Suspended load; Unmanned Aerial Vehicle (UAV); Unmanned aerial vehicles
• ISSN: 0957-4174; 1873-6793
• DOI: 10.1016/j.eswa.2021.115380

•UAV with load suspended by cable Intelligent control.•Quadrotor with cable-suspended load mathematical model by hybrid automata formalism.•NN continuously estimates external disturbance generated by suspended-load.•On-line learning of the neuro-estimator increases the robustness of the control.•Good UAV cyclic helical and polyline trajectory tracking. Unmanned aerial vehicles (UAVs) have been proved very useful in civil and military sectors: defense, security, shipping, construction, agriculture, entertainment, etc. Some of these applications, especially those related to transport and logistic operations, require the use of suspended loads that may make the vehicle unstable. In order to deal with this non-linear complex system with a changing mass, further research on modelling and control must be developed. In this work, a new intelligent control strategy is proposed and applied to a quadrotor with a cable-suspended load. The UAV carrying a suspended load has two different dynamic behaviors, depending on the state of the cable. Thus, we proposed to model the complete system using the hybrid automata formalism. Using this novel UAV model approach, a hybrid control is designed based on feedback linearization controllers combined with an artificial neural network, which acts as an online estimator of the unknown mass. The suspended load is dealt with as an external disturbance. Simulation results show how the on-line learning control scheme increases the robustness of the control and it is able to stabilize the quadrotor without any information about neither the position of the load nor the tension of the cable. Additionally, the computational complexity of the proposal is studied to show the feasibility of the implementation of this intelligent control strategy on real hardware.