Role-Based Mechanisms For Human-Computer Collabration In Systems Involving Humanwares

Human-computer collaboration has been used in various senarios. Human and computers have their own advantages, and do well in different tasks. Human-computer collaboration can enhance the performance of a collaborative system because the system involves the abilities of human and computers.Humanware is a novel technology that supports human-computer collaboration. Its core is an expert, aided by some necessary software and hardware interfaces. Because of these interfaces, the expert can fulfill his/her potential. In addition, humanwares, softwares, and hardwares can be assigned to different tasks dynamically. Systems involving humanwares have advantages of human and computers, yet encounter chanllenges. Agents’abilities for tasks should be evaluated before collaboration. Then each agent must be assigned to an appropriate task. Furthurmore, agents’ abilities changes as the collaboration progresses, thus task reassignment is indispensible.Role-based collaboration (RBC) is a methodology that paves the way for overcoming abovementioned challenges. It designs, runs and optimizes a system from the perspective of a group. With RBC. the common goal of the group is divided into roles (subtasks), and the main job of an agent is to play a role (perform a subtask). If only agents are evaluated, assigned, and reassigned properly, the group can obtain high performance.In terms of agent evaluation, an approach is proposed in response to the changing ability of an agent. This approach considers the agent’s abilities of autonomy and interaction, as well as the historical and current evaluation values. Experiment results indicate that the proposed approach outperforms the approach only utilizing the current evaluation value when the agent’s ability changes significantly.This dissertation discusses two kinds of role assignment problems:the role assignment problem with a flexible formation and the categorized bottleneck assignment problem. Their objective functions are to maximize the sum of chosen evaluation values and the performance of the role with the minimum performance. These two problems are both solved by linear programming. An adaptive collaboration (AC) approach is also presented to promote the optimal solution to the categorized bottleneck assignment problem. The effectiveness and efficiency of the proposed approaches are verified by experiments.The AC approach is stated in detail when solving role reassignment problem. Three algorithms are proposed for solving AC problems. They are based on three different scenarios:the current group state, the group state after a specific period, and the group state throughout the collaboration. More complex AC problems and their solutions are then investigated. Derived from the above-mentioned algorithms, two additional algorithms are presented. They consider reassignment costs. Experiments are developed to analyze the performance of each proposed algorithm. Results indicate that the proposed algorithms perform better than static collaboration where there are no reassignment costs. If reassignment costs exist, we also provide a way of determining whether to adopt an AC approach.This dissertation ends by implementing a humanware that can classify a picture. The accuracy of the humanware when classifying pictures is higher than software, and the workload of the humanware is reduced compared to only a person. A humanware service based on the service-oriented architecture is also developed. This humanware service can be published and called online.