基于IALO和BP神经网络的支气管镜机器人系统

Bronchoscopy Robot System Based on IALO and BP Neural Network

摘要: 支气管镜机器人在医生诊治传染病患者等需要远程治疗的手术中起到重要作用，传统支气管镜机器人PID控制效果较差，难以满足该手术主从同步控制的位置跟踪性能要求，针对这个问题新设计了支气管机器人系统和配套的主从式IALO-BP-PID机器人控制算法。BP-PID控制器通过自适应学习能力在设备响应过程中动态改变PID控制器参数让系统拥有更好的控制性能。但传统BP神经网络性能受网络参数初值影响较大，针对这个问题提出了基于改进蚁狮算法(IALO)优化权值的BP-PID控制方法，在传统蚁狮算法(ALO)中引入自适应系数和反调节因子改善算法寻优性能，在6个基准函数上的测试证明了改进算法的收敛与寻优能力优于原算法与其余对比优化算法。利用改进蚁狮算法(IALO)为BP神经网络选取参数初值配合神经网络的实时在线调整，实现了对神经网络控制器离线粗调和在线细调，提高了神经网络的性能。与传统PID控制算法的仿真相比，该控制算法的系统无超调且调节时间缩短了八分之七，使得支气管机器人的控制具有更佳的稳定性和跟随性，有一定的现实意义。

Abstract: Bronchoscopic robots play an important role in doctors’ diagnosis and treatment of infectious dis-ease patients and other surgeries that require remote treatment. Traditional bronchoscopic robots have poor PID control performance, making it difficult to meet the position tracking performance requirements of master-slave synchronous control in this surgery. To address this issue, a bronchial robot system and a supporting master-slave IALO-BP-PID robot control algorithm have been de-signed. BP-PID controller dynamically changes PID controller parameters during equipment re-sponse through Adaptive learning ability, so that the system has better control performance. How-ever, the performance of traditional BP neural networks is greatly affected by the initial values of network parameters. In response to this problem, a BP PID control method based on the improved ant lion algorithm (IALO) was proposed to optimize the weights. Adaptive coefficients and inverse adjustment factors were introduced into the traditional ant lion algorithm (ALO) to improve the op-timization performance of the algorithm. Tests on six benchmark functions have shown that the improved algorithm has better convergence and optimization capabilities than the original algo-rithm compared to other optimization algorithms. By using the improved ant lion algorithm (IALO) to select initial parameter values for the BP neural network and real-time online adjustment of the neural network, offline coarse and online fine tuning of the neural network model was achieved, improving the performance of the neural network. Compared with the simulation of traditional PID control algorithm, the system of this control algorithm has no overshoot and the adjustment time has been shortened by seven eighths, making the control of the bronchial robot have better stabil-ity and follow-up, which has certain practical significance.