Adaptive Feedforward Control Applied to Robotics

这项研究的目的是探索机器人如何通过练习轨迹来更准确地跟踪该轨迹。研究的重点是自适应前馈控制，其中前馈命令根据先前的性能误差进行修改。对所提出的轨迹学习算法进行数学分析、仿真评估，并在麻省理工学院串联式直接驱动机械臂上实现。这项研究的目标是让机器人，以及整个机器，移动得更准确。这项研究将提高机器人的性能，所开发的思想和算法可以直接应用于各种机器的控制。机器人和其他机器将能够以更快的速度更准确地移动，从而实现更快的吞吐量和更高的生产率。此外，更准确的性能将使对机器人进行编程以执行特定任务变得更容易，因为程序员将不再需要预测和补偿动态错误。

The goal of this research is to explore how robots might follow a trajectory more accurately with practice of that trajectory. The research focuses on an adaptive feedforward control, where a feedforward command is modified on the basis of previous performance errors. Proposed trajectory learning algorithms will be analyzed mathematically, evaluated through simulation, and implemented on the MIT Serial Link Direct Drive Robot Arm. The objective of this research is to make robots, and machines in general, move more accurately. This research will lead to improved robot performance, and the ideas and algorithms developed can be applied directly to the control of a wide variety of machines. Robots and other machines will be able to move more accurately at higher speeds, allowing faster throughput and greater productivity. In addition, more accurate performance will make it easier to program robots to do particular tasks, as programmers will no longer have to anticipate and compensate for dynamic errors.