基于前视节微电刺激的熊蜂机器人精确控制指令集研究

Micro electric stimulus is delivered into the neural systems and muscles of insects to direct their locomotion behavior, which is developed and referred as the cyborg insects. Characteristic of superior sensory and perceptual abilities, simplified control methods, low power and excellent concealment over man-made rivals, cyborg insects are regarded as one of the most competitive candidate of micro air vehicles and thus become the research focus in the multidisciplinary community. However, current cyborg insects are challenged with the roughly behavior inducing methods and also the heavy backpack to hinder the novel bio-machine hybrids from reaching their bright potential. To achieve the goal of control the insect precisely, this project will firstly investigate the mapping relationship between the micro electric stimuli and the responding behaviors of wing flapping and abdomen swinging of the bumblebees (for their advantages of superior locomotion, strong body and outstanding sensory abilities) with tethered conditions. Implemented using a lightweight backpack, the basic instruction set of initiation, cessation and asymmetric wing flapping control will then be developed as the micro electric stimulation patterns determined by the mapping relationship. Then, wireless energy transfer method is proposed to drive the backpack to produce stable micro electric stimuli and we will achieve free-flying cyborg bumblebees. Finally, graded control instruction set is explored by training the bumblebees to associate the graded stimuli with the turn angles based on their excellent associative learning ability. The precise control instruction set is composed by the graded instructions of start, stop, turn right, and turn left. The overall control effect of the instruction set will be verified by the navigation scenarios in a 3D sand table. This project aims to develop a novel computational model and key technical solutions for developing the precise control instructions for cyborg bumblebees, also beneficial to promote cyborg insects become really practical flying platforms.

昆虫机器人通过微电刺激直接控制昆虫的运动行为，具有感知运动能力强、控制系统简化、能耗低以及隐蔽性好等显著优点；因而成为了神经行为学、人工智能、信息电子等学科交叉研究的前沿热点。然而，现有昆虫机器人存在行为控制较为粗略、刺激背包过重等不足，无法实现自由状态下的精确控制，极大限制了其应用前景。本项目以运动和认知能力优异的熊蜂为载体，首先研究固定条件下微电刺激前视结节诱发行为响应的基本方法，建立“起始、中止、不对称振翅”等基本控制指令；提出无线、无源、轻量级刺激背包的设计策略，实现自由飞行的熊蜂机器人；进而，利用熊蜂联想学习能力，分级强化指令的控制效果，建立精确控制指令集，实现“起飞、中止、左、右转向”等分级控制，并在模拟复杂环境中测试指令集的行为控制效果。本项目研究将为熊蜂机器人的精确行为控制提供计算模型与关键解决方案，有助于推动昆虫机器人成为实用化的微型飞行平台。

昆虫机器人通过微电刺激直接控制昆虫的运动行为，具有感知运动能力强、控制系统简化、能耗低以及隐蔽性好等显著优点；因而成为了神经行为学、人工智能、信息电子等学科交叉研究的前沿热点。然而，现有昆虫机器人存在行为控制较为粗略、刺激背包过重等不足，无法实现自由状态下的精确控制，极大限制了其应用前景。本项目以运动和认知能力优异的熊蜂为载体，通过研究固定状态下微电刺激产生的行为响应关系和ES-BR神经计算模型，建立基本控制指令集；研究支持基本控制指令集的轻量级微电刺激背包，满足熊蜂负载约束，支持熊蜂长时间自由飞行；研究自由状态的熊蜂机器人行为训练范式，提出姿态估计方法，更新ES-BR计算模型，优化控制指令集；建立熊蜂机器人自动导航框架和系统，实现熊蜂机器人的自动导航原型。本项目研究将为熊蜂机器人的精确行为控制提供计算模型与关键解决方案，有助于推动昆虫机器人成为实用化的微型飞行平台。