|Thesis Advisor||谭民 ; 喻俊志|
|Place of Conferral||北京|
|Keyword||仿生滑翔机器海豚 多模态 深度控制 路径规划 路径跟踪|
The research on bio-inspired robotic fish and robotic dolphins aims to lay theoretical and technological foundations for design and control of high-performance underwater vehicles. There is still a large gap between robotic dolphins and underwater creatures in the aspects of motion performance and endurance, which deserves further investigation. This dissertation mainly concerns multi-modal modeling and motion control of a gliding robotic dolphin. The technical contributions are summarized as follows.
Firstly, in order to improve the endurance of the robotic dolphin, a hybrid scheme of the gliding robotic dolphin is proposed. By constructing its mechanical, hardware, and software systems, the prototype is developed. Considering that the gliding robotic dolphin can achieve both the gliding and dolphin-like motions, an integrated dynamic model is established based on Newton-Euler method, and the hydrodynamic forces are analyzed with the quasi-steady-state model. The simulations and aquatic experiments verify that the gliding robotic dolphin has good capabilities in three-dimensional~(3-D) multi-modal motion. Moreover, we conduct the experiments and simulations of the 3-D spiraling motion in the two modes, and the obtained results validate the effectiveness and accuracy of the established model.
Secondly, a depth control framework for the gliding and dolphin-like motions is proposed. In the aspect of gliding motion, a depth control method based on model predictive algorithm is proposed. By simplifying the dynamic model, a speed estimator based on sliding mode observer and a yaw controller based on PID are designed, further solving the problems of large delay and poor control accuracy. With regard to the dolphin-like motion, a dual-mode depth control method is particularly presented. The control method is composed of three main components: (1) The line-of-sight method is adopted to convert the depth control to pitch control. (2) Through combining with the characteristics of dolphin-like motion, a dual-mode selector based on flippers and CPG model is designed, and their basic principles are analyzed. (3) An adaptive method is employed to calculate the pitch control law, whose parameters are optimized offline based on the integrated dynamic model. Simulations and aquatic experiments demonstrate the effectiveness of the proposed methods.
Thirdly, a hierarchical framework to achieve path planning and following for a gliding robotic dolphin is proposed. Regarding the planer path planning as a multi-objective optimization problem, a hierarchical deep Q network is presented with the designed state space, action space, and reward function. By applying multiple small capacity networks to separately train the target approach path and obstacle avoidance path, the training speed is improved with satisfactory real-time performance. With regard to the path following, the smoothness problem in switching stage is addressed with an improved line-of-sight method. Further, we derive a nonlinear control law based on adaptive backstepping technique, and specially avoid singularities in the law derivation using Barrier Lyapunov function. Besides, the mapping laws based on piecewise and fuzzy method are presented to solve the control assignment problem of force/moment to actuator parameters. Both simulations and aquatic experiments are performed to verify the effectiveness of the proposed methods.
Fourthly, a 3-D path planning framework based on hybrid motion modes is proposed when the gliding robotic dolphin works in the marine environment. In order to enhance the yaw maneuverability, the multiple modes of yaw motion are particularly designed with the improved steering mechanism, and their performances are verified and analyzed by extensive aquatic experiments. Besides, a pitch maneuvering strategy based on a finite state machine is designed to realize the excellent vertical motion. Based on the motion characteristics of the 3-D maneuverability analysis, the feasibility of the planning task is verified by the dynamic model, and the kinematic constraints are obtained simultaneously. Further, the 3-D path planning method is composed of three main components, including the global path generation with gliding motion, obstacle avoidance path planning with dolphin-like motion, and path smoothing. Simulation results demonstrate the effectiveness of the proposed method.
|王健. 仿生滑翔机器海豚的多模态运动控制研究[D]. 北京. 中国科学院大学,2021.|
|Files in This Item:|
|Recommend this item|
|Export to Endnote|
|Similar articles in Google Scholar|
|Similar articles in Baidu academic|
|Similar articles in Bing Scholar|