基于视觉感知的海生物吸纳式水下机器人目标捕获控制

Vision Based Target Capture Control for Sea Organism Absorptive Underwater Vehicle

针对近海水产养殖环境下海生物目标的机器捕捞,设计了一款海生物吸纳式水下机器人.该水下机器人可以采取手动遥控吸取和视觉伺服控制吸取方式完成对海生物目标的机器捕捞.为了实现基于视觉感知的目标捕获控制,在摄像机平面坐标系上建立了水下机器人和目标之间的运动学关系,并在此基础上提出了自适应递归神经网络控制器.通过设计递归神经网络估计和补偿外界环境干扰,利用S面函数使水下机器人快速到达期望位置并保持稳定,结合递归神经网络和系统动力学模型设计鲁棒函数进一步提高非线性系统在视觉控制中的可靠性和稳定性.最后,在近海自然养殖条件下对海生物进行视觉跟踪控制实验,实现了对海生物目标的主动吸取控制,验证了该控制器的功能.

For the machinery capture of sea organism targets in the offshore aquaculture, a sea organism absorptive underwater vehicle is designed. The underwater vehicle can absorb and capture sea organism targets in manual tele-operation mode and visual servo mode. In order to realize vision based target capture control, the kinematic relationship between the underwater vehicle and the target is established in the camera plane coordinate, and the adaptive recurrent neural network controller is proposed on this basis. The recurrent neural network is designed to estimate and compensate environmental disturbance. The S surface function is utilized to make the underwater vehicle reach the expected position quickly and remain stable. Moreover, a robust function on the basis of recurrent neural network and system dynamic model is designed to improve the reliability and stability of the nonlinear system in visual control. Finally, a visual tracking control experiment for sea organisms is conducted in the offshore aquaculture, and active absorption control of sea organism targets is achieved,which verifies the controller function.