Controlling and programming parallel learning based on experiences in the intelligent system
Ye WeiQiong, Yongquan Yu · 2009
When the intelligent system takes experimental actions or its state transition, it will form control experiences impersonally. According to this fact, the paper proposes to acquire motion rule and programming rule under unsupervised condition from the experiences in parallel so as to form the control knowledge of this system. In the system study, the system deduces rules from experiences and further generalizes the rules to concepts or rules in higher level, constructing multiresolutional knowledge architecture. In the experiment, the mobile robot effectively learned and programmed the system's state, accomplished action control and achieved diversified goals in quasi-optimal manner under unsupervised condition on the basis of random experiences. At the same time, the system processes the ability to adapt itself to the new environment and mission.