Knowledge Petri Net and Logical Reasoning

Knowledge reasoning is a key area in artificial intelligence research, focusing on deriving unknown information based on existing knowledge within an environment, such as knowledge bases and inference rules. The environments in which intelligent agents operate are often partially observable and uncertain. Consequently, knowledge bases typically include not only deterministic knowledge but also uncertain knowledge, posing significant challenges to reasoning.

To address this, our team has proposed a probabilistic independent pruning algorithm for Knowledge Petri Nets. This algorithm’s network structure can simultaneously represent deterministic knowledge norms and prior probabilistic knowledge, substantially reducing computational complexity in uncertain reasoning and achieving exponential efficiency improvements. Building on this algorithm, we have developed a novel approach that enables unified representation and joint reasoning of deterministic and uncertain knowledge, enhancing flexibility in reasoning processes.

Moreover, our team has introduced a temporal uncertain knowledge reasoning method based on Petri Nets, which generates posterior probability distributions over infinite time steps within a finite temporal Knowledge Petri Net. This innovation provides a new solution to temporal reasoning challenges and effectively improves decision-making efficiency in uncertain environments.