毕业生知识库

  仿生机器鱼作为鱼类推进机理与机器人技术的融合，以其优异的性能被广泛应用于海洋勘测、军事侦察等诸多领域。随着近年来对机器鱼研究的持续深入，如何在实验水池环境下完成准确易用的机器鱼定位则显得尤为必要。传统机器人定位方法由于环境限制、适装性差、系统复杂等原因通常难以直接被应用于实验环境下的机器鱼定位上。本文以所设计的仿生机器鱼平台为定位载体，构建传感通信系统，考虑机器鱼运动特点及不同实验环境，针对包括短距离、中长距离以及不依赖外围传感设备在内的三组不同机器鱼定位场景基于多传感器融合技术设计了多种机器鱼定位方法并进行了实验验证与分析。本文主要工作如下：
  首先，本文设计了一种仿箱鲀机器鱼并构建了传感通信系统框架作为定位方法研究与验证平台。本文对包括方位、视觉、信号、压力等在内的多种传感器进行了选型配置，并对主控电路及基本模块进行介绍。
  其次，本文针对不同场景及机器鱼运动特征实现多种机器鱼定位方法：在近距离下，本文基于多模态传感数据进行对机器鱼方位进行融合解算从而在实验水箱环境中实现机器鱼定位；在中长距离下，本文首先利用 BLE(Bluetooth Low Energy)单信号，设计基于 K-Means 的 CGC(Clustering Grid Correct)聚类网格矫正算法对四点定位坐标进行校准；其次本文采用 UWB-BLE(Ultra Wide Band - Bluetooth Low Energy)双信号实现定位，通过提出的基于动态权值模糊推理的自适应融合定位算法，进一步提高了基于信号的机器鱼定位方法的稳定性。本文还采用水压传感来对机器鱼不依赖外围传感设备的自定位进行探索。通过结合 Softmax 和交叉熵构建具有分化特性的速度级别预测 BP 神经网络并融合多水压传感及方位信息完成了对机器鱼在直游及类直游运动下的鱼游轨迹估测。
  最后，本文对研究工作进行了总结，对所提出的多种机器鱼定位方法进行评价探讨，并对包括侧线感知、自学习智能机器鱼定位等未来研究方向进行展望。

  As a fusion of fish propulsion mechanism and robotic technology, bionic robotic fish is widely used in marine survey, military reconnaissance and many other fields due  to  its  excellent  performance.  With  the  continuous deepening  of  robotic  fish research  in  recent  years,  it  is  particularly necessary  to  complete  accurate  and easy-to-use robotic fish positioning in the experimental pool environment. Traditional robot positioning methods are usually difficult to be directly applied to robotic fish positioning  in experimental  environments  due  to  environmental  limitations,  poor adaptability, and complex systems. Taking the designed bionic robot fish platform as the  positioning  carrier,  this  work constructs  the  sensor communication  system,  considers  the  motion  characteristics  of  robot  fish  and  different  experimental environments,  and  designs  a  variety  of  robot  fish  positioning  methods  based  on multi-sensor fusion technology for three groups of different robot fish positioning scenes  including  short  distance,  medium-to-long  distance  and  independent  of peripheral sensing devices.
  First,  this  work designs  an ostraction-like  fish  and  builds  a sensor communication  system  framework  as  a  research  and  verification platform  for positioning methods. This paper selects and configures a variety of sensors including orientation, vision, signal, pressure, etc., and introduces the main control circuit and basic modules.
  Secondly, this work implements a variety of robotic fish localization methods for different scenarios and robotic fish motion characteristics:
  At a short distance range, this work fuses the orientation of the robotic fish based on multi-modal sensor data to realize robotic fish positioning in the experimental water tank environment; at a medium-to-long distance, firstly, using BLE((Bluetooth Low Energy) single type signal, this work designs a CGC clustering grid correction algorithm  based  on  K-means  to  calibrate  the  four  point  positioning  coordinates; secondly, this paper uses UWB-BLE(Ultra Wide Band - Bluetooth Low Energy) dual type signals to realize positioning. Through the proposed adaptive fusion positioning algorithm based on dynamic weight fuzzy reasoning, the stability of signal-based robot fish positioning method is further improved. This work also employs water pressure sensing to explore the self-localization of robotic fish independent of peripheral sensing devices. By combining Softmax and cross  entropy  to construct  a  speed  level  prediction  BP  neural  network  with differentiated  characteristics,  and  integrating  multiple  water  pressure  sensors  and orientation information, the fish swimming trajectory estimation of the robotic fish under straight and straight-like motions is completed.
  Finally, this work summarizes the research work, evaluates and discusses the various  proposed  robotic  fish  localization  methods,  and  looks forward  to  future research  directions  including  lateral  line  perception  and  self-learning  intelligent robotic fish positioning.