Reinforcement Learning In Control Systems For Walking Hexapod Robots
Ivo Hrdlička, Patrik Kutílek
Abstract
We have developed a controller for a six-legged robot that allows it to walk on rough terrain or where mines have been detected. Intention of this article is to describe application of reinforcement learning methods for implementation of an acyclic walking. Cyclic methods of a walking are not feasible in difficult terrain because doesn’t consider subsequent undercarriage states of the robot in future. For basic (not so rough) terrain we used new designed heuristic algorithm. In complex environment, computational more intensive approach (i.e. Q-learning) is applied. Overall control algorithm of the walking undercarriage combines these two presented techniques. Results of the new algorithm of control to the problem of learning to walk with a six legged robot are presented by demining robot.
Full paper 2005-09-10 17:09
The Locomotion Control of the Concyclical Walking Robot
Patrik Kutílek, Jiří Kacer
Abstract
The basic aim of the paper is to propose the solution of the concyclically six-legged walking carriage intended for the military purposes to walking in the highly broken topography. On the locomotion level are solved – with help of the intelligent control methods – both movement of the legs of the carriage and evaluation of the sensorial information from the sensorial subsystem of the robot. For fulfilling of these requirements are used methods of the machine learning without priory information about the conflict environment.
Full paper 2005-01-10 17:06
Contemporary Trends in Development and Modernization of Rocket Systems
Milan Machala, Radek Doskočil
Abstract
This article deals with contemporary trends in development and modernization of rocket systems of field artillery. Especially reasons for modernization of rocket systems are mentioned. These reasons emerge from nature of contemporary armed conflict in the world. The main attribute of modernization is implementation of control system on the board of rockets.
Full paper 2005-01-03 17:12