2024-03-29T04:54:14Z
https://repository.dl.itc.u-tokyo.ac.jp/oai
oai:repository.dl.itc.u-tokyo.ac.jp:00005669
2022-12-19T03:47:06Z
6:209:271
9:233:234
2ノルムに基づく冗長性解消法性能向上の研究 : ロボット制御への適用
Improved 2-norm-based Redundancy Resolution Methods : With Application to Robotics
Baratcart, Travis
11785
修士(工学)
Redundancy is a useful characteristic in a multitude of systems and is valued for its ability to endow dexterity and fault tolerance in systems operating in dynamic, remote, or unpredictable environments. Although useful, incorporation of redundancy in a system endows added complication in the control structure which needs to be resolved. // Easily the most popular method of resolving this redundancy is through application of the 2-norm optimizing pseudoinverse. Resolution through 2-norm optimization is very popularly utilized due to its analytical tractability: the 2-norm resolves systems uniquely, continuously, and often most importantly, in a very simple way. A well known problem exists in 2-norm resolution, however, in that it fails to make use of systems’ full potential output space. Due to the complexity of alternatives, system designers tend to either ignore this problem and work within the bounds of 2-norm resolution or make use of methods to extend the resolution range of 2-norm resolution. // The most popular extension of the 2-norm is found in the Cascaded Generalized Inverse (CGI). CGI is an intuitive extension of 2-norm and prior to the work of this thesis held its place as the largest extension of 2-norm resolution. // In this work, three methods will be introduced based upon 2-norm resolution. The first two methods are modifications of CGI. It will be shown that although successful in extending the resolution range of 2-norm, CGI loses out on one of the main benefits of 2-norm resolution: continuity. Additionally, despite this drawback, CGI still does not attain the full output range of arbitrary systems. // The first proposed method, the Continuous Cascaded Generalized inverse (cCGI), is targeted at the first problem. cCGI is introduced as the largest extension of the 2-norm that ensures continuity of resolution. The continuity of CGI is analytically proven and the dynamic improvements when using cCGI, with respect to 2-norm and CGI, are simulated. // For systems in which this continuity is not a significant concern, or for such systems that have been sufficiently tested to ensure discontinuity does not arise, the Extended CGI (eCGI) is proposed. eCGI resolution is currently the largest extension of 2-norm resolution. // Both of these system are introduced and compared with application to kinematic redundancy in robotic manipulators. // Finally we look at a particular system, biarticular actuation redundancy, which is unique in that there exist multiple resolution schemes with the simplicity of 2-norm in implementation. The presence of these resolution solutions has allowed us to propose the first realization of 2-norm/Infinity-norm switching resolution. These two norms are physically preferable in opposite circumstances, and connecting the two allows for greater utility than either method used alone. The continuity of this switching system is analytically proven, and the system is experimentally implemented on a robot arm. It is shown that utilization of switching resolution improves both the motor size (with respect to 2-norm) and energy requirements (with respect to infinity-norm) of the system.
thesis
2014-09-26
2014-09-26
application/pdf
https://repository.dl.itc.u-tokyo.ac.jp/record/5669/files/37126941.pdf
eng