D.V. Blinov – Student, Saint-Petersburg State University of Aerospace Instrumentation
O.Yu. Sivchenko – Programmer, St. Petersburg Institute for Informatics and Automation of RAS
A.R. Shabanova – Junior Research Scientist, St. Petersburg Institute for Informatics and Automation of RAS
Reverse kinematics problem (RKP) is one of the key problems in the robotics domain. The solution of this problem ensures planning of local movement and robotic device control. But, concerning multi-joint robotic systems, it is often complicated or impossible to obtain RKP analytical solution in an explicit form. A further complication arises in connection with existence of multiple solutions because of kinematic ambiguity of complex systems.
The most popular solution approaches for this problem are analytical and geometrical ones. But, for structures, consisting of five joints or more (in some configurations the number of joints in this context is limited to 3), RKP is impossible to solve via classic methods. Hence it becomes necessary to employ various workarounds, allowing to bypass limitations of the classic methods.
The most problems, associated with classic solution methods, can be rectified using machine learning approaches. In this paper one of such methods is considered, specifically a genetic algorithm.
Genetic algorithms belong to bioinspired machine learning approaches and perform well during offline optimization. This advantage provides for genetic algorithm usage in various robotic contexts: development of local control systems for robot actuators, tracing op-timal route for robotic device as such.
The objective of this work is to develop a compositional RKP solution method for a four-joint manipulator based on genetic algorithms. Two genetic algorithms were designed to solve this task: the first for searching manipulator joint angles with known target point coordinates and the second for searching manipulator joint angles with known target point coordinates and previously found joint angles for other known target points in proximity. We also proposed a composition method for sequential application of the developed genetic algorithms, which ensures 99% accuracy in reaching the target point by the terminal manipulator link, as well allows reducing RKP solution time by 70…95% compared to basic genetic algorithm approach. The RKP solution approach showed decent efficiency in tests on PhantomXPincher manipulator model in V-REP environment. Because of its flexibility, the proposed algorithm applies well to other manipulator types with rotational joints. Also, due high computational speed of the algorithm it works well with physical models. Additional advantages of the proposed genetic algorithms compared to other machine learning algorithms in RKP context are that the algorithm developed here requires no pre-training. So, its application seems reasonable, when it is necessary to solve RKP for a limited point set.
- Medvedev S.N., Yakovlev A.Yu., Korolkov O.G. Razrabotka metoda opredeleniya prisoedinennykh parametrov manipulyatora s vrashchayushchimisya zvenyami na osnove geneticheskikh algoritmov. Vestnik Voronezhskogo gosudarstvennogo tekhnicheskogo universiteta. 2018. T. 14. № 6. S. 8−15. (In Russian).
- Momani S., Abo-Hammour Z.S., Alsmadi O.M.K. Solution of inverse kinematics problem using genetic algorithms. Applied Mathematics & Information Sciences. 2016. V. 10. № 1. P. 225.
- Köker R., Çakar T. A neuro-genetic-simulated annealing approach to the inverse kinematics solution of robots: a simulation based study. Engineering with Computers. 2016. V. 32. № 4. P. 553−565.
- Zhou Z., Guo H., Wang Y., Zhu Z., Wu J., Liu X. Inverse kinematics solution for robotic manipulator based on extreme learning machine and sequential mutation genetic algorithm. International Journal of Advanced Robotic Systems. 2018. V. 15. № 4.
- Starke S., Hendrich N., Magg S., Zhang J. An efficient hybridization of genetic algorithms and particle swarm optimization for inverse kinematics. IEEE International Conference on Robotics and Biomimetics (ROBIO). IEEE. 2016. P. 1782−1789.
- Starke S. A Hybrid Genetic Swarm Algorithm for Interactive Inverse Kinematics. Universität Hamburg, Fachbereich Informatik. 2016.
- Galemov R.T., Ivannikov I.V. Planirovanie traektorii manipulyatora dlya dvizhushcheisya tseli. Kibernetika i programmirovanie. 2018. № 2. S. 9−28. (In Russian).
- Bakdi A., Hentout A., Boutami H., Maoudj A., Hachour O., Bouzouia B. Optimal path planning and execution for mobile robots using genetic algorithm and adaptive fuzzy-logic control. Robotics and Autonomous Systems. 2017. V. 89. P. 95−109.
- Ryadchikov I.V., Gusev A.A., Sechenev S.I., Nikulchev E.V. Geneticheskii algoritm poiska parametrov PID-regulyatorov sistemy stabilizatsii shagayushchego robota. Trudy NGTU im. R.E. Alekseeva. 2019. № 1(124). (In Russian).
- Datta R., Pradhan S., Bhattacharya B. Analysis and design optimization of a robotic gripper using multiobjective genetic algorithm. IEEE Transactions on Systems, Man, and Cybernetics: Systems. 2015. V. 46. № 1. P. 16−26.
- Lyu V. Metody planirovaniya puti v srede s prepyatstviyami (obzor). Matematika i matematicheskoe modelirovanie. 2018. № 1. (In Russian).