Architectures for integrating manufacturing activities and enterprises : T.J. Williams, P. Bernus, J. Brosvic, D. Cheng, G. Doumeingts, L. Nemes, J.L. Nevins, B. Vallespir, J. Vlietstra, D. Zoetekouw, and other members of the IFAC/IFIP Task Force, pp 883–893
- Publisher
- Elsevier Science
- Year
- 1995
- Tongue
- English
- Weight
- 136 KB
- Volume
- 3
- Category
- Article
- ISSN
- 0967-0661
No coin nor oath required. For personal study only.
✦ Synopsis
For a pendnlum-cart-DC-motor system, an on-line self-learning control scheme has been based on a reward/punishment approach. Unlike in previous work, the goal is simply to get the pendulum from its hanging position into its upright position, with no further requirements concerning the motion. No iuformation about the equation of motion is required; the reinforcement signal of the learning scheme is calculated from basic physical measurements. Learning is both structural and parametric. The learning system determines how many parameters have to be used, and to which values they must be set. These parameters constitute the memory of the control algorithm. Computer simulations are presented.