Research Initiation: Neural Network Robot Controller

首席研究员(PI)将研究人工神经网络在机器人控制中的使用。针对单连杆机械手对各种对象施力的情况，对各种非线性自适应控制问题进行了仿真研究。柔性和刚性机械手将与移动和固定对象相互作用，执行PI指定的各种任务。非线性自适应控制器将结合学习机械手的非线性动力学的人工神经网络。计算机模拟将被用来评估基于神经网络的机器人控制器的性能。在前馈控制器结构中，将利用神经网络逼近代表机器人机械手动力学的多变量非线性函数的能力。研究重点将集中在用于刚性和柔性单链系统的神经网络的学习算法上。此外，还将考察控制器对磨损或拾取物体引起的动态变化的适应性，以及控制器的替代架构。这个项目的目标是创建一种适用于任何控制问题的神经网络控制器。

The principal investigator (PI) will study the use of artificial neural networks in robotic control. A variety of nonlinear adaptive control problems will be simulated for a one-link manipulator applying force to various objects. Flexible and rigid manipulators will interact with mobile and fixed objects to conduct various tasks specified by the PI The nonlinear adaptive controller will incorporate an artificial neural network that learns the nonlinear dynamics of the manipulator. Computer simulations will be used to assess the performance of robotic controllers based on neural networks. The ability of neural networks to approximate multivariate nonlinear functions representing robot manipulator dynamics will be exploited in a feedforward controller architecture. Research will focus on learning algorithms for neural networks used with both rigid and flexible one-link systems. Adaptation to varying dynamics caused by wear or by picking up objects will also be examined, as will alternative architectures for controller. the goal of this project is to create a neural network controller that is applicable to any control problem.