基于复合传感器的机器鱼推进力预测方法研究

Research on Propulsion Prediction Method of Robotic Fish Based on Composite Sensor

摘要:

复杂多变的水下环境会极大地干扰仿生机器鱼的正常运动和对水下信息的感知，而鱼鳍的变形在产生推进力方面发挥了关键作用。但目前对于鱼鳍如何利用自身感觉信息来对推进力进行识别和预测的研究甚少。为了解决这一问题，通过在一个柔性传感器上集成压电层和压阻层，研制出基于压电/压阻双传感模式的柔性仿鱼鳍复合传感器，并提出了一种基于复合传感器对鱼鳍产生的推进力预测的方法。建立了基于 BP 人工神经网络的推进力预测模型并进行训练，通过皮尔逊相关分析法对输入层数据进行筛选和优化，提高了模型的预测精度。实验结果证明基于 BP 神经网络及皮尔逊相关性分析的多传感层组合的推进力预测是有效可行的，同时为机器鱼水下感知能力的提升提供了新的途径。

Abstract:

The complex and variable underwater environment can greatly interfere with the normal movement and perception of underwater information of bionic robotic fish，and the deformation of fins plays a key role in generating propulsive force. However，little research has been conducted on how fins use their sensory information to identify and predict propulsive force. In order to solve this problem，a flexible fish fin-like composite sensor based on piezoelectric/piezoresistive dual sensing mode is developed by integrating a piezoelectric and piezoresistive layer on a flexible sensor. A method is proposed to predict the propulsive force generated by the composite sensor. A propulsive force prediction model is established and trained，which is based on BP artificial neural network. The prediction accuracy of the model is improved by screening and optimizing the input layer date through Pearson correlation analysis. The experimental results demonstrate that the propulsive force prediction which is based on combination of multiple sensing layers of BP neural network and Pearson correlation analysis is effective and feasible，and also provides a new way to improve the underwater sensing capability of the robotic fish.