自然界中的绝大部分生物均处于流体环境中，其自主推进运动通常伴随着主动与被动变形，例如鸟类的飞翔和鱼类的游动等。为了适应外界环境的变化，这些生物逐步进化出了多样的高性能推进方式。另一方面，部分生物为了完成捕食、洄游等生存任务，在不断探索中掌握了关键的生存策略。生物独特且高效的推进方式和控制自身适应环境的决策手段是各类仿生自主推进器设计的最直接灵感来源。为了探究生物自主推进运动背后的流体力学效应，本文利用分布式力矩驱动模仿鱼类的肌肉驱动，研究了仿生柔性扑翼的长程直线巡游、近壁面推进和转向机动三种游动方式，并分析了驱动参数和雷诺数对其推进性能的影响。此外，利用深度强化学习方法，研究了仿生柔性扑翼在复杂环境中的自适应行为。本文的主要研究成果和创新点如下：

针对主动和被动综合驱动下仿生柔性扑翼的自主推进问题，采用流固耦合的数值模拟方法对其游动性能进行了系统的测试。主动驱动采用了一个行波形式的弯矩来模拟分布式肌肉驱动，被动驱动则通过流固耦合考虑。探讨了波数、矩密度幅值和雷诺数等关键参数对长程直线游动性能的影响。通过模拟贴近壁面时的游动状态，研究了地面效应对同周期运动形态、巡航速度和推进效率的影响，并对近壁时的偏移现象做出了物理解释。此外，设计了一种简单的前缘偏转角演化规律来完成转向机动行为，实现了多方向稳定偏航。本研究结果有望为仿生自主水下航行器的设计提供参考。

针对仿生柔性扑翼在复杂环境中的自适应行为控制问题，设计了一种深度强化学习-流固耦合(DRL-FSI)实时通讯交互求解框架。基于深度递归Q网络(DRQN)的强化学习算法，通过构建代理模型的方式，对游泳者在均匀来流中的趋流性行为和静态流中的点对点游动行为等自适应行为进行了数值模拟，同时指出了目前计算中存在的问题并提出了改进措施。本研究结果可以为仿生装置的控制提供一定的理论指导。
 

Most of the creatures in nature are immersed in fluid environment, and their self-propelled locomotion is usually accompanied by active and passive deformation, such as the flight of birds and swimming of fish. To adapt to the changes in the environment, these creatures have evolved a variety of high-performance propulsion strategies. On the other hand, in order to complete the survival tasks such as predation and migration, some of the creatures have mastered the key survival strategies in constant exploration. The unique and efficient propulsion mode of creatures and the decision-making means of controlling their own adaptation to the environment are the most direct inspiration for the design of various bio-inspired autonomous propulsors. In order to explore the hydrodynamic effects behind the self-propelled locomotion of creatures, three swimming modes of the bio-inspired flexible flapping wing, including the long-range straight-line swimming, the near-wall propulsion and the turning maneuver, are studied by using distributed moment actuation to imitate the muscle actuation of fish, and the influence of driving parameters and Reynolds number on its propulsion performance is analyzed. In addition, the adaptive behavior of bio-inspired flexible flapping wing in complex environment is studied by using deep reinforcement learning. The main research achievements and innovative works of this dissertation are as follows:

Aiming at the problem of self-propelled locomotion of bio-inspired flexible flapping wing under active and passive actuation, fluid-structure interaction numerical simulation method is used to systematically test its swimming performance. The active actuation adopts moment in the form of a traveling wave to mimicking distributed muscle actuation, while the passive actuation is considered by fluid-structure interaction. The influences of some key parameters, such as the wavenumber, the amplitude of moment density and the Reynolds number, on the performance of straight-line swimming are explored. The influence of the ground effect on the co-periodic motion pattern, cruise speed and propulsion efficiency are investigated through the simulation of near-wall swimming, and the physical explanation of the migration phenomenon near the wall is given. In addition, a simple evolution law for the leading-edge deflection angle is designed to realize the turning maneuver behavior, and the multi-direction steady yaw is realized. The results of the present study are expected to be helpful to the design of bio-inspired autonomous underwater vehicles.

Aiming at the adaptive behavior control problem of bio-inspired flexible flapping wing in complex environment, a deep reinforcement learning-fluid-structure interaction (DRL-FSI) real-time communication interactive solution framework is designed. Based on the deep recurrent Q-network (DRQN) reinforcement learning algorithm, by constructing a proxy model, the adaptive behaviors of swimmer's rheotaxis in uniform flow and point-to-point swimming in static flow are simulated numerically. At the same time, the existing problems in the calculation are pointed out and the improvement measures are put forward. The results of the present study can provide some theoretical guidance for the control of bio-inspired devices.