|Place of Conferral||北京|
|Keyword||机器人操作规划 全局柔顺性 高精度装配 柔顺抓取|
How to achieve high precision and compliance for robotic systems has been a long-term concern in the field of robotics research. In the intersecting fields of robotics and (i) control science, (ii) mechanism science, (iii) material science, and (iv) computer science, the question above has been studied from extensive angles. However, how to make the robots realize human-level or human-like dexterity and compliance is still a focused issue in related fields. In addition, it is one of the key problems that restrict the further application of robots in the industry of high-precision essential component manufacturing, 3C manufacturing and so on.
In this paper, attempts will be started from the point of view of simulating humans, to create an overall framework for robotic compliant manipulation, through analysis of the motion structure and decision-making methods of human in the process of manipulation, with fully consideration of the integration of environmental constraints and sensing information, which leads to forming new robotic compliance algorithms and corresponding hardware platform to:
1) fulfill the requirement for typical robotic manipulation tasks,
2) improve the generalization ability of the algorithms for robotic manipulation tasks, and
3) reduce the dependence of system on the precision of the sensing information, and improve the stability of the system through the utilization and integration of environmental constraints and sensing information.
It is hoped that the work could be helpful for the further application of robots in the manufacturing and service industries in China.
On the basis of the group's research on attractive region in environment (ARIE) for sensor-less high-precision manipulation, and aiming at the above objectives, this thesis will study the conditions of existence and availability of environmental constraints in high-dimensional configuration space. The integration of incomplete sensing information and environmental constraints will also be considered, and the strategy for high-precision robotic manipulation will be designed.
The main research contents are as follows:
First, a new high-precision manipulation strategy planning method is proposed, utilizing high-dimensional environmental constraints. Aiming at the problem of high-precision peg-hole assembly for 3D convex parts, the mapping relationship between the contact states in the configuration space and the geometric shape of the parts in the physical space is analyzed. The conditions of the existence of available attractive region in environment (ARIE) are also given. It is proved that high dimensional ARIE widely exists in the assembly process of 3D convex parts. By analyzing the configuration space of a class of cylindrical parts with convex polygon section, the general ARIE-based strategy for high-precision 3D convex parts assembly is presented.
Second, the framework and strategy planning method for robotic compliant manipulation based on the integration of high-dimensional environmental constraints and sensing information are proposed. In view of the general process of robotic manipulation, the expression of sensing information in non-physical space is studied based on the analysis of the high-dimensional environmental constraints formed in the configuration space between the robot and the object to be operated. The stability conditions of the system state in the integrated space of environmental constraint and sensing information are proved, and the strategy planning based on integration of the environment constraint and sensing information is designed in the new space. Further, targeting at the hand-eye system with force/torque feedback, a compliant grasping method is proposed to achieve rapid response to catch fast moving objects, which proves the effectiveness of this strategy.
Third, an overall framework and method for realizing compliance in robotic manipulation systems are proposed. Through analyzing human physiological structure, the hardware basis of achieving compliance in human manipulation is discussed; Based on human behavioral experiments, the action pattern and decision making mechanism during human manipulation is explored, which could been recognized as the software basis of achieving compliance in human manipulation. Combining the above analysis, some human-inspired motion structure and mechanism are introduced to robot systems, in hope that through the integration of sensing information and environmental constraints, a robot-applicable framework for achieving overall compliance like human, is proposed. The framework could be used to realize the effective adjustment to the process of compliant manipulation.
Fourth, based on the proposed method, the compliant robotic assembly system with high precision is designed, which utilizes the integration of high-dimensional environmental constraints and sensor information. Through effective fusion of ``machine-constraint-information'', the assembly of essential components for RV reducer and some typical parts are taken as the typical case, to illustrate the validity of the manipulation strategy, where the position and pose of the parts are quickly located by using the low-precision vision and the grip-part constraint and the matching error are eliminated by using the low precision force / torque information and the matching constraint between the parts.
Finally, the research results are concluded, where a discussion on how the integration of environmental constraints and sensing information helps in achieving high precision, high compliance and high generalization as the same time is also provided. Discussions on future work are also given.
|李睿. 基于高维环境约束与不完备传感信息融合的机器人高精度柔顺性操作研究[D]. 北京. 中国科学院大学,2018.|
|Files in This Item:|
|基于高维环境约束与不完备传感信息融合的机（21662KB）||学位论文||暂不开放||CC BY-NC-SA||Application Full Text|
|Recommend this item|
|Export to Endnote|
|Similar articles in Google Scholar|
|Similar articles in Baidu academic|
|Similar articles in Bing Scholar|
Items in the repository are protected by copyright, with all rights reserved, unless otherwise indicated.