Development of an artificial neural network topology for generating the motion of robotic manipulator
Motion planning is an important issue in robot industry. Without an appropriate motion planning, a robot may be colliding with obstacles or passing through undesirable points. In order to control the motion of a robot manipulator, a person has to possess the knowledge of kinematics, dynamics, and tr...
Saved in:
| Main Author: | |
|---|---|
| Format: | Thesis |
| Language: | English |
| Published: |
2014
|
| Subjects: | |
| Online Access: | http://psasir.upm.edu.my/id/eprint/47951/1/FK%202014%204R.pdf |
| Tags: |
Add Tag
No Tags, Be the first to tag this record!
|
| Summary: | Motion planning is an important issue in robot industry. Without an appropriate motion planning, a robot may be colliding with obstacles or passing through undesirable points. In order to control the motion of a robot manipulator, a person has to possess the knowledge of kinematics, dynamics, and trajectory planning.
However, there are two main problems in using conventional methods. Firstly, the equations are hard to be derived and the calculations are complex. Secondly, the characteristics of different trajectories are different and there is no mathematical solution for unknown trajectory. Hence, the first objective in this research is to simplify the complex calculations in terms of solving kinematics and trajectory
planning issues simultaneously. Another objective of this research was to help in computing the motion of a manipulator even though the characteristic of the
trajectory is unknown. In order to achieve these research goals, artificial neural network (ANN) was proposed as a solution.
In the early stage, a virtual manipulator was developed and subjected to different primitive trajectories. In order to examine the ability of ANN in tracking the motion of a robot manipulator, a primitive ANN would be used to track the moving path of the virtual robot manipulator’s end effector in the virtual environment. This ANN was developed based on the fundamental of back-propagation neural network
(BPNN) topology. The topology of ANN would be modified for reducing the errors and deviations. Eventually, the developed ANN would be validated through a real time 5 catalyst robot. Besides, obstacle avoidance planning would be integrated into the developed ANN. Virtual obstacles would be allocated within the robot’s workspace randomly and the performances of developed ANN would be observed
through simulation experiments.
The results indicated that ANN possessed ability in tracking the motion of a robot manipulator in terms of solving kinematics and trajectory planning issues
simultaneously and it was able to compute the motion of a manipulator even though the characteristic of the trajectory was unknown. Obstacle avoidance planning was
integrated into the architecture of developed ANN for better performances and the results were satisfactory. With this developed method, a person is able to compute a
safe path for a robot manipulator to avoid obstacles (objects which enclosed in a sphere). |
|---|