Intelligent Motion Planning And Flight Control For Rotorcraft Unmanned Aerial Vehicle Transportation Systems

Rotorcraft unmanned aerial vehicles(UAVs)are widely used in transportation,mapping,monitoring and other fields due to their simple structure,strong maneuverability,high safety,and flexible control,which are gradually developing towards the direction of autonomy and intelligence.In particular,in order to improve the efficiency,reliability,and intelligence of transporting goods,the motion planning and flight control problems of the rotorcraft UAV transportation system have received much attention from industry and academia.Specifically,motion planning generates safe and efficient desired trajectories for the system which takes into account real-time performance and optimality,and flight control ensures that the system can accurately track the planned desired trajectories.Considering the complex external environment,the nonlinear and state-coupling characteristics of the system,designing intelligent motion planning and flight control strategies to achieve efficient and reliable air transportation,has very important theoretical significance and broad application prospects.Although researchers at home and abroad have carried out in-depth theoretical research on the motion planning and flight control of the rotorcraft UAV transportation system and achieved certain results,the following issues remain to be resolved from the perspective of practical application: 1)In unknown environments,based on real-time information of sensors,it is difficult for existing motion planning strategies to generate accurate intelligent decisions online,resulting in a lack of intelligent planning methods that take into account robustness and reliability.2)In terms of trajectory planning along a given path,it is difficult for the existing methods to guarantee the security,real-time implementation and optimality of the algorithm simultaneously,and the model-based planning scheme is difficult to apply to the actual environment with uncertainties.3)Regarding the rigid connected payload,it is difficult for existing flight control methods to guarantee the tracking performance theoretically,and most existing intelligent control methods lack the necessary generalization ability for practical applications.4)For the cable-connected payload,most flight control methods do not fully utilize the coupling characteristics of the payload,the cable and the rotorcraft UAV,and the control accuracy and adaptability still need to be improved.To address the above-mentioned practical problems,in-depth research on motion planning and flight control problems of the rotorcraft UAV transportation system is conducted in this paper.The main contributions are summarized as follows.(1)To solve the motion planning problem in unknown environments,a novel real-time trajectory generation framework is proposed utilizing the data-driven high-level decision making and model-driven low-level optimization,which guarantees reliable and intelligent navigation of the rotorcraft UAV transportation system in unknown environments.Specifically,high-level reinforcement learning strategy generates accurate intelligent decisions rapidly,such as faster,slower,etc.,based on real-time observations from sensors.According to the decision and model information,the low-level optimization strategy computes the motion-level trajectories in real time.In addition,by introducing expert strategy and imitation learning in the training of the high-level strategy,the reliability and generalization ability of the decision making are improved.Benefiting from designing the learning and optimization strategies in the framework,the scheme enhances the intelligence of the system and realizes real-time motion planning that takes into account optimality,robustness and reliability.Finally,compared with the existing methods,the high efficiency and strong reliability of the proposed algorithm are verified by the experimental results.(2)Regarding the time-optimal trajectory planning problem along a given path,the differential flatness property of the rotorcraft UAV transportation system is first deduced,then the states are projected into the path coordinate space,and by designing slack variables and nonlinear transformations,the state constraints are transformed into convex constraints.On this basis,the input part of the planning solution is intelligently adjusted by the designed reinforcement learning strategy,which suppresses the influence of uncertainties,such as model errors and external disturbances,in an online interactive way,and improves the accuracy and reliability of the transportation along a given path.Different from the existing planning strategies,the reinforcement learning is introduced based on the convex optimization,which intelligently plans the actual input in an interactive way,showing stronger robustness and adaptability.A series of comparative experiments are conducted to verify the high path following accuracy and robustness of the proposed method.(3)For the flight control problem of the rigid connected payload,two effective robust control methods are proposed in this paper.For one thing,by utilizing the disturbance observer to estimate the uncertainties in the system,a nonlinear robust control method is proposed to improve the robustness of the control system effectively.Then,a nonlinear feedback control strategy is designed based on barrier Lyapunov function,which successfully restricts the tracking error in the given safe range.For anther,in order to further improve the intelligence of the control system,by combining the robust integral of the signum of the error(RISE)method with the data-driven reinforcement learning,a novel intelligent control framework is proposed in the paper.Specifically,an asymptotically convergent RISE control strategy is designed firstly,which guides the update direction of the reinforcement learning strategy.Then,the accurate judgment of the pros and cons of explorations is achieved based on the dual-critic training framework,then the reinforcement learning strategy is updated with remarkable good exploration actions,ensuring reliable and efficient policy convergence.Several comparative experiments are conducted to verify the excellent performance of the proposed control methods in terms of control accuracy and robustness.(4)Aiming at solving the flight control problem of the cable-connected payload,the regulation control and trajectory tracking control methods are proposed in the paper.First,the coupling characteristics between the payload,the cable and the rotorcraft UAV are fully analyzed,and an open-loop error system is established based on the introduced auxiliary variables and virtual control inputs.On this basis,an intelligent regulation control method is proposed to achieve accurate point-to-point payload transportation.Meanwhile,the control parameters of the payload are generated intelligently by the reinforcement learning strategy,which ensures the positioning accuracy and robustness of the system.Then,for trajectory tracking tasks,a nonlinear tracking control method is proposed,where the corresponding auto-tuning laws of the payload and cable control parameters,are designed to achieve accurate trajectory tracking control.The control parameters of the system can be automatically adjusted to appropriate values in different tasks,which reduces the dependence on parameter tuning and improves the adaptability of the control scheme.Finally,theoretical analysis and practical experiments are conducted to fully verify the effectiveness of the proposed regulation control and trajectory tracking control algorithms.