Dynamic Control of Intention Priorities of Human-Like Agents

1 School of Computer Engineering, Nanyang Technological University, Singapore 639798, email: {pg04043187, ASSYHUANG}@ntu.edu.sg Abstract. Intention scheduling mechanism plays a critical role in the correct behaviour of BDI agents. For human-like agents, the independent intentions should be scheduled based on their priorities, which show the respective importance and urgency of the intentions. We propose to enrich the BDI agent architecture with 2 processing components, a PCF (Priority Changing Function) Selector and a Priority Controller. These enable a BDI agent to assign an initial priority value to an intention and change it with time according to the chosen PCF. The initial priority value reflects its urgency at the time of intention creation. The PCF selected defines how the priority should change with time. As an example, we design a function by simulating human behaviors when dealing with several things at the same time. The priority first increases with time according to the Gaussian function to simulate that people are more inclined to do something which has been on their mind for sometime. After a certain time, if the intention still does not get executed because of other higher priority intentions, its priority will decrease according to the Ebbinghaus forgetting curve. Experiment results show that with this mechanism, the agent can show some human-like characteristics when scheduling intention to execute. This can be used when simulating human-like agents.