Collaborative Research: Decentralized Adaptive and Extremum Seeking Control of Robot Manipulators Using Image Processing

大型机电代理网络，包括复杂的机器人机械手和自动驾驶汽车，很快就会变得无处不在。这类系统的先进控制将越来越多地受益于廉价的固态摄像头，但由此产生的基于视觉的系统将是计算密集型的，并带来显著的延迟。众所周知，延迟会导致控制系统不稳定，可能会导致关键基础设施出现灾难性故障。该项目将开发方法来补偿和克服网络控制系统中的这种延迟。该项目还将有助于科技人力资源的开发，继续调查人员的努力，他们在过去的项目中监督了20多名来自代表性不足群体的学生，包括妇女、西班牙裔和非裔美国人。该项目开发的方法是为有效使用图像数据而量身定做的，以促进实时多代理路径规划和碰撞避免。这将使用几何表示法和基于特征的学习相结合来实现。方向和3-D位置将针对HSV颜色分类的要抓住的对象、障碍物和放置对象的目的地进行表征。这些方位和3-D位置将被用于实时生成用作所需轨迹的无碰撞路径。将制定计算高效、完全分散和节能的极值搜索控制方案，利用固有的并行和高度冗余的处理体系结构，并在存在相互关联的非线性、饱和和不确定的情况下运行。将使用28阶Baxter机器人对结果进行实验验证，以检查算法的健壮性和计算能效。该奖项反映了NSF的法定使命，并通过使用基金会的智力优势和更广泛的影响审查标准进行评估，被认为值得支持。

Large networks of electromechanical agents, including complex robotic manipulators and self-driving cars, will soon become ubiquitous. Advanced control of such systems will increasingly benefit from inexpensive solid-state cameras but the resulting vision-based systems will be computation-intensive and introduce significant delays. Delays are known to cause instability in control systems, leading potentially to catastrophic failures in critical infrastructure. This project will develop methods to compensate and overcome such delays in networked control systems. The project will also contribute to the development of human resources in science and technology, continuing the efforts of the investigators who have supervised two dozen students from underrepresented groups on past projects, including women, Hispanics, and African-Americans. The methods developed in the project are tailored to the efficient use of image data to facilitate real-time multi-agent path planning and collision avoidance. This will be accomplished using a combination of geometric representations and feature-based learning. Orientations and 3-D positions will be characterized for HSV color-classified objects to be grasped, for obstacles, and for destinations where the objects are to be placed. These orientations and 3-D positions will be employed for real-time generation of collision-free paths to be used as desirable trajectories. Computationally efficient fully decentralized and energy-efficient extremum seeking control schemes will be formulated, leveraging inherently parallel and highly redundant processing architecture and operating in the presence of interconnected nonlinearities, saturations, and uncertainties. Experimental validation of the results, using a 28th-order Baxter robot, will be carried out to examine the robustness and to validate the computational-energy efficiency of the algorithms.This award reflects NSF's statutory mission and has been deemed worthy of support through evaluation using the Foundation's intellectual merit and broader impacts review criteria.