|
Mobile robot motion planner via neural network
Krejsa, Jiří ; Věchet, Stanislav
Motion planning is essential for mobile robot successful navigation. There are many algorithms for motion planning under various constraints. However, in some cases the human can still do a better job, therefore it would be advantageous to create a planner based on data gathered from the robot simulation when humans do the planning. The paper presents the method of using the neural network to transfer the previously gained knowledge into the machine learning based planner. In particular the neural network task is to mimic the planner based on finite state machine. The tests proved that neural network can successfully learn to navigate in constrained environment.
|
|
Experimental Determination of Human Arm Force/Elbow Joint Angle Relation
Krejsa, Jiří ; Zezula, M. ; Věchet, Stanislav
The achievable force at the end of human forearm depends on the forearm position. The actual value of this force is one of the most important variables affecting the design of the motorized splint, but reliable data are hard to find. Sophisticated measuring device was constructed to determine this force. The force is measured during the movement of the forearm, as such method yields better results than static measurement. The trajectory of movement is enforced by the measuring device, so the influence of the subject on the trajectory is eliminated.
|
|
Rychlá metoda srovnávání skenů pro lokalizaci mobilního robotu
Krejsa, Jiří ; Věchet, Stanislav
The paper is focused on fast method which uses subsequent proximity sensor scans of mobile robot environment to determine the robot position and orientation and to build a local map. The environment is supposed to be unknown. The method, based on Potential-Based Scan Matching which gives a measure of scan match is combined with gradient descent in order to find the proper match of two scans in reasonable time. Tests of robustness against the noise in sensor readings are included.
|
|
Nadřízený software pro autonomního závodního robota
Krejsa, Jiří ; Věchet, Stanislav ; Ondroušek, V.
The paper describes the high level software issues in the task of autonomous robot driving on the park pavement. The task in question is to use the sensor data in order to autonomously navigate on the path given in predefined map. The paper gives detail information on the development of software tools for the task and our experience with different ways of sensors data acquisition.
|
| |
| |
| |
|
Markovská lokalizace pro mobilní roboty: simulace a experiment
Krejsa, Jiří ; Věchet, S.
Summary: Localization of the robot is a task of estimating robot position in known environment from sensor observation. The paper describes basic principles of Markov localization technique, succesfully used for localization task. Method is robust against sensor errors and can deal with global uncertainty when robot position is completely unknown. Both simulation and experimental verification of method usability are included.
|
|
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.
|
| |