基于动态避障风险区域的仿生机器鱼路径规划方法

A Biomimetic Robotic Fish Path Planning Method Based on Dynamic Obstacle-avoidance Risk Region

针对来流速度固定、存在多个静态漂浮障碍物和动态障碍物的复杂水环境中胸尾鳍协同推进仿生机器鱼的自主避碰问题,提出了一种基于动态避障风险区域(DAR)的路径规划方法。首先,结合所建立的机器鱼水动力学模型,利用扩展卡尔曼滤波法构造沿障碍物运动方向的类椭球形动态避障风险区域,其长轴与障碍物的运动速度成正比,并通过模糊控制方法对卡尔曼滤波过程的噪声方差进行估计,得到该区域的精确边界;其次,结合机器鱼视场内障碍物的位姿与速度,去除同向运动且速度大于机器鱼的无风险障碍物,实时获得动态环境中的避障风险区域集合,进而得到机器鱼可通行时变区域;最后,根据先避近后避多、边界距离最短原则初步确定先转向、后俯仰的空间避碰策略,进而以障碍物作为外界扰动,设计以期望位姿为输入的非线性模型预测控制器,实时优化得到机器鱼的转弯半径、俯仰角和两侧胸鳍相位差等控制参数,驱动机器鱼安全、快速地通过当前障碍物区域。实验结果表明,机器鱼通过多障碍物区域时,与风险区域边界的最小距离为0.15 m,速度最高达0.15 m/s,空间避障速度最高为0.3 m/s,运动机动性较高且运动轨迹比较平滑,验证了所提方法的有效性。

Aiming at the autonomous collision-avoidance problem of the pectoral and caudalfin co-propelled biomimetic roboticfish in a complex aquatic environment with afixed incomingflow velocity and the presence of multiple staticfloating obstacles and dynamic obstacles,a path planning method based on dynamic avoidance risk region(DAR)is proposed.Firstly,an ellipsoid-like dynamic avoidance risk region along the direction of obstacle movement is constructed according to the established hydrodynamic model of the roboticfish by using extended Kalmanfiltering,whose long axis is proportional to the speed of obstacle movement,and the noise variance of the Kalmanfiltering process is estimated through a fuzzy control method,to get the precise boundary of the region.Secondly,the non-risky obstacles moving in the same direction as the roboticfish at a greater speed are removed according to the position and speed of obstacles in thefield-of-view of the roboticfish,the set of obstacle avoidance risk regions in the dynamic environment is obtained in real time,and then the time-varying passable region of roboticfish is acquired.Finally,a spatial collision-avoidance strategy is preliminarily determined,in which the roboticfish steersfirstly and then pitches according to the principles of avoiding the nearest obstaclefirstly and then plenty of other obstacles,and keeping the shortest distance from the safe region boundaries.Taking the obstacles as an external perturbation,a nonlinear model predictive controller is designed with the desired pose as input to optimize the control parameters such as turning radius,pitch angle and phase difference between two pectoralfins in real time,so as to drive the roboticfish to pass through the current obstacle area safely and quickly.The experimental results show that when the roboticfish passes through the multi-obstacle area,the minimum distance from the boundary of the risk area is 0.15 m,the speed is up to 0.15 m/s,the spatial obstacle avoidance speed is up to 0.3 m/s,and the spatial maneuverability is high and the trajectory is relatively smooth,which verifies the effectiveness of the proposed method.