Learning a visual task by genetic programming

Chongistitvatana, P.; Polvichai, Jumpol

doi:10.1109/iros.1996.570847

Cited by 3 publications

(2 citation statements)

References 6 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…In this approach, the aim was to evolve the correct inverse kinematics equations and the values for each of the joints' parameters simultaneously. A method based on genetic programming was also employed to control a robotic manipulator with 3 DOF to reach a target and avoid obstacles (Chongstitvatana & Polvichai, 1996). A look up table was used instead of kinematics equations.…”

Section: Illmentioning

confidence: 99%

“…Therefore, the learning capabilities of the system are limited (Koza, 1992). A learning system is defined as one that can adapt its structure over time in order to improve its performance (Chongstitvatana & Polvichai, 1996). One of the current challenges is that the learning time must be limited in robotics, since some controllers can take hundreds of thousands of trials to evolve (Gu et al, 2003, Mataric & Cliff, 1996.…”

Section: Introductionmentioning

confidence: 99%

See 1 more Smart Citation

Genetic Programming Control of an Articulated Robotic Manipulator

Ward¹,

Siddique

McGinnity³

et al. 2008

Journal of Intelligent Systems

View full text Add to dashboard Cite

The inverse kinematics problem is one of the most important and challenging within articulated robotic manipulator scenarios. Classical mathematical techniques in solving the solutions to robotic kinematics control problems involve meticulous and time-consuming operations. There are no general methods in solving inverse kinematics problems or good practical solutions. Computer aided modeling methods have been developed in the absence of such straightforward numerical solutions, however they can also be computationally expensive. Recently, genetic programming approaches have been studied in robotic manipulator kinematics problems, where the inverse kinematic equations and joint variables are evolved simultaneously. An approach based on the genetic programming paradigm where the joint variables are evolved and used with an existing forward kinematics model is proposed and is used to control an articulated robotic manipulator with 5 degrees of freedom. Results show a successful controller can be evolved, where a number of optimal solutions are discovered initially and the solution that involves the least energy cost is selected.

show abstract

Section: Illmentioning

confidence: 99%