Over-subscribed problems are commonly encountered in both industry and in our daily lives. From planning a weekend trip to managing a large factory floor, we often need to deal with inconsistent objectives and/or incomplete planning. To efficiently find a good resolution, it is necessary to combine the knowledge and experience of humans and the reasoning capabilities of autonomous systems.
This is the motivation for my thesis research and the development of Uhura, an interactive decision support system. It is designed to support a human-robot team in resolving over-subscribed planning problems. The core of Uhura is a conflict-directed diagnosis algorithm that can detect conflicts between planning goals and incomplete planning domains. To address the conflicts and incompleteness, the algorithm needs to generate alternative plans by relaxing some of the goals and domains. Once a feasible alternative is found, Uhura will initialize a dialog with the user to explain the cause of failure and present the alternative. The user can either take Uhura’s advice, or ask Uhura to find different alternatives. This failure-driven dialog is initialized each time one or more conflicts are detected that threaten the user’s plan, during both planning and execution phases.
Uhura greatly reduces the effort required to resolve large-scale over-subscribed problems, and enables a more fluent human-robot collaboration in planning. Through the interaction, it also improves the quality of the solutions by continuously getting user inputs, and adjusting to any changes in the situation. Currently, I am exploring the applications of Uhura in three fields:
1. Transportation: a personal transit assistant for optimizing commute trips, and a robust route management tool for service providers.
2. Deep-sea Explorations: a mission advisor for planning and scheduling activities with large uncertainty and managing available resources.
3. Manufacturing: a factory floor planning advisor for repairing broken plans due to unexpected failures.
You can download my PhD Thesis Proposal from here. You may find a list of my research and course projects on the Projects and Seminars page. Let me know if you want to learn more about them and I am happy to share more information.