Research and Development of Visual Perception and Manipulation in an Interactive Autonomous Robotic System

该计划的拟议提案是开发方法和技术，以在具有动态交互的实际应用中实现对自主移动机器人的精确操纵。主要贡献是提高目前自主机器人系统的通用性、自主性和可靠性，特别是在操纵应用方面。考虑到最终目标所需的能力，该计划的范围涵盖四个主要研究领域：1.机器视觉，特别是实时和稳健的目标跟踪和手势理解；2.先进的力、触觉和视觉控制；3.机器学习；4.传感器融合。提出了系统的总体架构，包括三个子系统：1)。高效的视觉跟踪和运动估计；2)。交互框架；3)。先进的操纵控制。 在所提出的体系结构中，高效的视觉跟踪为基于视觉的控制提供实时且鲁棒的目标跟踪和运动估计结果。提出的交互框架包含一个描述任务的知识库和一个高级解释器，它可以为给定任务中的姿势、力和触觉信息提供参考。此外，还将开发一种先进的控制系统，它将根据高级任务解译人员建议的姿势实现手相对于对象的对准，以及一个位置-力-触觉控制器，以在接触过程中提供稳健和稳定的操作。 该计划将通过开发具有交互框架的先进的位置-力-触觉控制器，为自主机器人系统的操作、运动和运动估计的高效光学跟踪以及类似人类的操作带来创新的交互框架。该项目的研究成果将在搜救、家庭护理和辅助机器人等许多实用领域具有巨大的应用潜力。开发的智能自主机器人适用于工业和日常生活，提高了工作效率，减轻了人类的工作量，并有效地执行了需要重型机械的任务。这些发展的技术有望在机器人研究方面取得重大进展，特别是在感知、交互和操作方面，并为自主机器人提供新的研究方向。

The proposed proposal for the program is to develop methodologies and technologies that will lead to precise manipulation of autonomous mobile robots in practical applications with dynamic interaction. The main contribution is set to improve the versatility, autonomy and dependability of current autonomous robotic systems, specifically in manipulation applications. In view of the needed capabilities for the ultimate objective, the scope of the proposed program spans four main research areas: 1. Machine vision, specifically real-time and robust object tracking and hand gesture understanding 2. Advanced force, tactile and vision control, 3. Machine learning, and 4. Sensor fusion. An overall architecture is proposed, containing three sub-systems: 1). Efficient visual tracking and motion estimation; 2). Interaction framework; and 3). Advanced control in manipulation. In the proposed architecture, the efficient visual tracking provides real-time and robust object tracking and motion estimation results to the vision-based control. The proposed interaction framework contains a knowledge base that describes tasks, and a high-level interpreter that can provide references for pose, force and tactile information in a given task. Also, an advanced control system will be developed, which will achieve the alignment of hand with respect to the object according to the suggested pose by the high-level task interpreter, and a position-force-tactile controller to provide robust and stable manipulation during the contact. The proposed program will lead to innovative interaction framework for an autonomous robotic system in manipulation, efficient optical tracking in motion and motion estimation, and human-like-manipulation by developing an advanced position-force-tactile controller with an interaction framework. The research results in the proposed program will have significant potential for application in many practical areas including search and rescue, homecare and assistive robots. The developed intelligent autonomous robots are applicable in industry and in daily life, leading to improved work efficiency, reduced workload for humans, and effective execution of tasks that require heavy duty machinery. The developed technologies are expected to result in significant progress in robotics research, particularly concerning perception, interaction and manipulation, and provide new research directions in autonomous robotics.