One of the central problems in Artificial Intelligence is to generate a plan (i.e., a sequence of actions) to fulfill the agent's goal. High-level program execution is an approach to this problem, and it is based on the situation calculus. It requires the user to encode his comprehension about the domain into a control program that contains nondeterministic choices, and the interpreter transforms this high-level program to a feasible plan.
If the program is almost deterministic, a plan can be produced very efficiently.

Otherwise, the complexity of high-level program execution is more like that of planning, which is another solution to this problem. By and large, high-level program execution is more efficient than planning. Currently, the area of high-level program execution mainly concerns single-agent settings .However, in most scenarios many agents interact with each other. Each agent decides the next action according to her mental attitudes not only about the world but also about other agents' mental attitudes.

The papers about this project are as follows:
  1. Liangda Fang and Yongmei Liu, "Multiagent Knowledge and Belief Change in the Situation Calculus", In Proceedings of the Twenty-Seventh AAAI Conference on Artificial Intelligence (AAAI-2013), pages 304-312, 2013.
  2. Liangda Fang, "High-Level Program Execution in Multi-Agent Settings", In Proceedings of the Twenty-Third International Joint Conference on Artificial Intelligence Doctoral Consortium Program (IJCAI-2013 DC), pages 3213-3214, 2013.

Last edited Apr 20, 2014 at 6:01 AM by comun, version 7