• Home
  • Current congress
  • Public Website
  • My papers
  • root
  • browse
  • IAC-05
  • C1
  • 4
  • paper

    • Optimal trajectory of robots using symbolic regression

    Paper number

    IAC-05-C1.4.07

    Author

    Ms. Zuzana Oplatková, Tomas Bata University in Zlín, Czech Republic, Czech Republic

    Year

    2005

    Abstract
    The contribution deals with finding an optimal trajectory for a robot using a new tool for symbolic regression – Analytic Programming (AP).  To control movements of robots is necessary in space industry because of preciously realization of the required tasks as soon as possible. Problem of this kind was examined also by John Koza in his Genetic Programming which is similar tool as Analytic Programming. It is so called Multiagent problem. The aim of this problem is to find an optimal trajectory of agents (e.g. robots) (shortest way and shortest time) to perform the required tasks. 
The tool used for simulations is Analytic Programming which is a new tool for symbolic regression from the field of artificial intelligence. AP is a super-structure of evolutionary algorithms.  It is a universal tool for creation a complex formula from simple functions. In the case of fitting unknown measured data the simple functions are e.g. plus, minus, variables, constants and the complex formula is e.g. (3x + 4) / x. They can be also linguistic terms as it is in the case of this article e. g. turn right, turn left, move towards, pick something up. Analytic Programming serves to connect them together to the complex formula – description of all journey of robot here. 
The rest is up to evolutionary algorithms which are used for optimization problems. The journey of the robot can be laid down as minimization of the Hamming distance between just synthesized and required behaviour. In the contribution two newer evolutionary algorithms – SelfOrganizing Migrating Algorithm (SOMA) and Differential Evolution (DE) are used. They were successfully used in many difficult optimization tasks.
SOMA is an evolutionary algorithm classified as algorithm with competitive – cooperative strategy. It means that new population in evolution is not produced by crossover of two or more individuals but individuals migrate on the surface of cost function and find the best solution (minimum). 
DE is, on the other hand, from group of algorithms with crossover based partially on genetic algorithms but used 4 parents to produce offspring for new population.
Analytic Programming is the similar tool as Genetic Programming in the sense of output but the structure is completely different. Results obtained during this simulation of this theoretical application according to author’s opinion might be used also in practical application in the space industry. 
Analytic Programming is new but studies and comparative studies with Genetic Programming were successfully published on conferences, e.g. ICICIS’02,Cairo, Egypt, Ciras’03, Singapore, SCI 2004, Orlando, USA.
    Abstract document

    IAC-05-C1.4.07.pdf

    Manuscript document

    IAC-05-C1.4.07.pdf (🔒 authorized access only).

    To get the manuscript, please contact IAF Secretariat.