Automation in Handling Uncertainty in Semantic-knowledge based Robotic Task-planning by Using Markov Logic Networks

Abstract

Generating plans in real world environments by mobile robot planner is a challenging task due to the uncertainty and environment dynamics. Therefore, task-planning should take in its consideration these issues when generating plans. Semantic knowledge domain has been proposed as a source of information for deriving implicit information and generating semantic plans. This paper extends the Semantic-Knowledge Based (SKB) plan generation to take into account the uncertainty in existing of objects, with their types and properties, and proposes a new approach to construct plans based on probabilistic values which are derived from Markov Logic Networks (MLN). An MLN module is established for probabilistic learning and inferencing together with semantic information to provide a basis for plausible learning and reasoning services in supporting of robot task-planning. In addition, an algorithm has been devised to construct MLN from semantic knowledge. By providing a means of modeling uncertainty in system architecture, task-planning serves as a supporting tool for robotic applications that can benefit from probabilistic inference within a semantic domain. This approach is illustrated using test scenarios run in a domestic environment using a mobile robot.