Neuronal path planning and motion control of mobile robots: F. Pourboghrat and M. R. Sayeh. Department of Electrical Engineering, Southern Illinois University, Carbondale, IL 62901-6603 USA
- Publisher
- Elsevier Science
- Year
- 1988
- Tongue
- English
- Weight
- 98 KB
- Volume
- 1
- Category
- Article
- ISSN
- 0893-6080
No coin nor oath required. For personal study only.
✦ Synopsis
Most industrial robots today are placed on fixed locations. A few have limited mobility on tracks mounted on the factory floor [I]. There are also mobile carts that transport workpieces, but these carts can move only in a structured environment, i.e., by following painted lines. Robot mobility, however, is needed in a wide variety of robot functions in unstructured environments such as mining, military operations, and aid to the handicapped. A major research issue in robot mobility is autmmmous control, which includes motor control, sensing, navigation, commtmication, obstacle avoidance, and task performance.
To generate a path for a mobile robot ls to find a continuous trajectory starting from the initial position of the robot to lts target position, in an environment which contains obstacles. The path generation should be based on the information about the geometric characteristics of the environment available to the robot from the pest or obtained directly from sensors. One such dynamic path generation is discussed in [2].
The usual approach to building control systems for such mobile robots is the hierarchical control, i.e., to decompose the problem into a series of functional units. A different approach to building the autonomous mobile robot control system is to use task achieving behaviors as the prim~ decomposition of the problem [3]. Such a controller architecture is simple, robust N easily modifiable. However, because of tba huge amount of information processing and computational burden, the real implon~etton of such controllers was not quite successful.
Recently, we have seen an explosion of interest and research in the field of neural networks. Noural networks have demonstrated powerful computational abilities through the use of massive, interconnected architectures which more closely resemble the structure of the brain than the traditional digital computer architecture. Networks composed of a large number of simple processing units with a high degree of connectivity can be trained to perform many types of pattern recognition and associative recall. Hopfietd and Tank [4] showed that many computotionally difficult problems can be formulated as minimization problems, which in turn can be easily solved using the neural networks.
In this paper, the problem of dynamic path generation and motion control of mobile robots is addressed. The path planning method of [2] and the layered control methodology of [3] are formulated as an optimization problem, and a neural network approach for the intelligant control of mobile robots is suggested. The design enables the robot to move from any initial position to any given target location, in an unstructured environment, avoiding unknown obstacles. Simulation results show the feasability of the method and the improved performance of the robot.
[1]Nitzan, D., "Development of Intelligent Robots: Achievements and Issues," IEEE J.