Analysis, Detection Policy, and Performance Measures of Detecting Task Planning Errors and Conflicts

Executive Summary In this research, we analyze the impact and severity of conflicts and errors that occur during task planning through collaboration. Taxonomy of conflict and error problems and alternative detection methods are categorized to review the different types of conflict and error situations. Three techniques are developed and applied for the impact analysis. First, a performance measure, " detect-ability " , is defined to evaluate the detection potential of given detection methods. This measure and its analysis can provide the necessary decision-making information for the design of conflict/error detection. In addition, a cost-damage analysis based on available conflict/error knowledge is also applied to help decide on the optimum detection policy. Third, a viability measure is adapted for detection protocols and agents, and is applied as a measure of impact on the company to examine the detection policy and detection performance for collaborative task planners. The paper describes these three techniques, and experimental results in using them. The experiments indicate clearly that a detection policy associated with the cost-damage analysis method can maintain significantly higher viability of the collaborating planners, hence, positive total impact. This result is important, because it points to specific design guidelines for CEDM (Conflict and Error Detection Management) systems.