Dual-arm robotic manupulation inspired by human skills

Abstract

Man’s desire to shape the world to fit his ideas (or vision) has led to major technological achievements. Many machines have been inspired by biological systems and processes observed in nature. The greatest challenge for the research engineer is to develop a human-like mechanical system that has the skills of a human being. The shape of the human body has largely inspired the mechanical structure of humanoid robots. Current efforts of researchers in humanoid robotics and cognitive sciences are aimed at translating models of human behavior to humanoid robots. Modeling of human movement has been extensively studied and explored in the literature, to design and control a humanoid robot inspired by human movement in daily human activities. Various scientific disciplines analyze human movement in different ways. There are many techniques and strategies for imitating and analyzing human motions, such as the motion imitation process, imitation learning, and optimization approaches. The objectives of this thesis are to: – improve the conversion process for the imitation of human upper-body movement by a humanoid, for a task with and/or without contact between hands and equipment; – define an objective function optimized by human movement, in order to transfer human skills to humanoid robots; and – execute humanoid robot motion similar to human motion, by minimizing the same criteria. Since the human model is complex and has many degrees of freedom, the goal is to imitate human motion and transfer human skills to a humanoid robot with fewer degrees of freedom. To this end, a kinematic model of the robot ROMEO was used as the kinematic model of a human body. The motions of a human being that uses both arms to perform a task, including contact with the environment, are analyzed in the thesis. This type of motion has not been sufficiently explored in the literature. A conversion process for the imitation of human motion by a humanoid robot is proposed in the first part of the thesis. This conversion process enhances existing techniques and was developed to imitate human motion and have a humanoid robot perform a task with and/or without contact between hands and equipment...

Cˇ ovekova želja da stvori svet po svojoj zamisli (ili viziji) dovela je do razvoja velikih tehniˇckih dostignu´ca. Inspiracije za stvaranje brojnih mašina istraživaˇci su pronašli posmatranjem bioloških sistema i procesa u prirodi. Najve´ci izazov za istraživaˇca je razvoj mehaniˇckog sistema nalik ˇcoveku koji ima karakteristike ljudskog bi´ca. Mehaniˇcka struktura ljudskog tela u velikoj meri je mapirana u model humanoidnih robota. Sadašnji napori istraživaˇca u oblasti humanoidne robotike i kognitivnih nauke su da transformišu modele ljudskog ponašanja i procesa u model humanoidnih robota. Modeliranje ljudskog pokreta široko je prouˇcavano i istraživano u literaturi sa ciljem da se dizajnira humanoidni robot i definiše njegovo upravljanje inspirisano karakteristikama ljudskog pokreta u svakodnevnim ljudskim aktivnostima. Razne discipline nauke analiziraju ljudsko kretanje na razliˇcite naˇcine. Postoje brojne tehnike i strategije koje imitiraju i analiziraju ljudsko kretanje, kao što su imitacioni proces, uˇcenje iz imitacije i optimizacioni pristupi. Cilj ove teze je: – da unapredi konverzione procese za imitaciju pokreta gornjeg dela tela ˇcoveka sa humanoidnim robotom za zadatke sa ili bez kontakata izmed¯u ruku i opreme; – da definiše objektivnu funkciju koja predstavlja optimizirano ljudsko kretanje sa ciljem prenošenja ljudskih veština na humanoidne robote; i – da humanoidni robot izvršava pokret na isti naˇcin kao ˇcovek minimiziranjen istih kriterijuma S obzirom da je model ˇcoveka kompleksan i obuhvata mnogo stepeni slobode, naš cilj je da se ljudski pokret i veštine kretanja prenesu na kretanje humanoidnog robota sa manje stepeni slobode. U tu svrhu, kinematski model robota ROMEO je koriš´cen za predstavljanje kinematskog model ljudskog tela. Pokreti dvoruˇcne manipulacije koji ukljuˇcuju kontakt sa okruženjem su predmet istraživanja u tezi. Ova vrsta kretanja nije dovoljno istražena u literaturi. U prvom delu teze predložen je konverzioni proces za imitaciju ljudskog kretanja sa humanoidnim robotom. Ovaj konverzioni proces poboljšava postoje´ce tehnike i razvijen je za imitaciju ljudskog kretanja sa humanoidnim robotom za izvod¯enje zadataka sa ili/i bez kontakta izmed¯u ruku i opreme...