|
Plánování cesty pro čtyřnohého kráčejícího robota použitím rychlých náhodných stromů
Krejsa, Jiří ; Věchet, S.
Summary: There are several randomized methods for problem of path planning. Rapidly exploring random trees (RRT) is a method which can deal with constraints typical for legged walking robots, e.g. limitations in rotation step resolution. Paper describes the RRT method itself and its use for path planning of four-legged walking robot, including special failure case when robot is capable of only rotating in one direction. The method proved to be robust and fast.
|
| |
| |
| |
|
Stochastic policy in Q-lerning used for control of AMB
Březina, Tomáš ; Krejsa, Jiří ; Věchet, S.
A great intention is lately focused on Reinforcement Learning (RL) methods. The article is focused on improving model free RL method known as Q-learning used on active magnetic bearing model. Stochastic strategy and adaptive integration step increased the speed of learning approximately hundred times. Impossibility of using proposed improvement online is the only drawback, however it might be used for pretraining and further fined online.
|
| |
|
Aplikace metody spojitého Q-učení
Věchet, S. ; Krejsa, Jiří ; Míček, P.
Standard algorithm of Q-Learning is limited by discrete states and actions and Q-functionis usually represented as discrete table. To avoid this obstacle and extendthe use of Q-learning for continuous states and actions the algorithm must bemodified and such modification is presented in the paper. Straightforward way isto replace discrete table with suitable approximator.
|
|
Design of small laboratory quadruped robot
Švehlák, M. ; Grepl, Robert ; Věchet, S. ; Bezdíček, M. ; Chmelíček, J.
In the article is a sum dectripcion of the design of a small quadrupedal walking robot. The aim of the work is to make a physical model for authenticating computational simulation and other problems related with walking robots. The physical model should be noted for simple design, unpretentious production and relatively small financial coast. The physical model is qualified enough to make various experiments. The actuators are concepted in reference to the character of the loading moment.
|
|
Aproximace modelu stability kráčivého robotu
Krejsa, Jiří ; Grepl, Robert ; Věchet, S.
The paper compares global and local approximation methods used for walking robot stability model. Global approximators are represented by feedforward multilayer neural network (FFNN) trained by gradient method; local approximators are represented by Locally Weighted Regression (LWR) and Receptive Field Weighted Regression (RFWR) methods.
|
|
Control of experimental valking robot using simulating model
Grepl, Robert ; Věchet, S. ; Bezdíček, M. ; Švehlák, M. ; Chmelíček, J.
This paper describes the scheme of remote operator control of mobile walking robot. Control structure is appropriate for low level control of robot (full operator supervision) and for initial phase of robot design. Kinematic and dynamic model are built in Matlab--Simulink--SimMechanics environment. Computation speed of model allowed real--time control of robot. Adopted approach is tested on small experimental physical model - quadruped walking robot with four legs and 12 d.o.f. overall.
|