FIELD: physics.
SUBSTANCE: invention relates to intelligent controllers using the principle of reinforcement learning, artificial neural networks, and can be used to control complex systems in a non-deterministic environment. Modified intelligent controller with adaptive critic comprises: reinforcement calculation unit, time difference error calculation unit, situation quality value delay unit, situation quality value multiplication unit, time difference error accumulator, critic unit, action selection unit, random number generator, neural networks training unit, operating parameter values error counter unit, decision neural network, operating parameter value subtraction unit, operating parameter value delay unit, action selection unit, action unit, control object.
EFFECT: high speed characteristics of the control system based on an intelligent controller, shorter training time for neural networks and simplification of the final implementation for the developer.
1 cl, 1 dwg
| Title | Year | Author | Number |
|---|---|---|---|
| MODIFIED INTELLIGENT CONTROLLER WITH ADAPTIVE CRITIC | 2013 |
|
RU2523218C1 |
| MODIFIED INTELLIGENT CONTROLLER WITH ADAPTIVE CRITICAL ELEMENT | 2020 |
|
RU2755339C1 |
| MODIFIED INTELLIGENT CONTROLLER WITH ADAPTIVE CRITIC | 2011 |
|
RU2450336C1 |
| MODIFIED INTELLIGENT CONTROLLER | 2011 |
|
RU2458390C1 |
| MODIFIED INTELLIGENT CONTROLLER | 2013 |
|
RU2527212C1 |
| INTELLIGENT CONTROLLER WITH SELF-MODIFICATION RULES OF EDUCATIONAL AND CONTROL NEURAL NETWORKS | 2011 |
|
RU2447494C1 |
| METHOD FOR CREATING CONTROLLERS FOR CONTROLLING WALKING ROBOTS BASED ON REINFORCEMENT LEARNING | 2022 |
|
RU2816639C1 |
| INTELLECTUAL CONTROLLER WITH NEURON NETWORK AND SELF-MODIFICATION RULES | 2003 |
|
RU2266558C2 |
| MODIFIED FUZZY RULE INTELLIGENT CONTROLLER | 2012 |
|
RU2504002C1 |
| CONTINUOUS CONTROL BY DEEP LEARNING AND REINFORCEMENT | 2016 |
|
RU2686030C1 |
