PROJECT-BASED ACTIVITY SCHEDULING FOR PERSON AGENTS

摘要

This is the third is a series of papers that are attempting to cumulatively develop an integrated conceptual framework for modelling both short- and long-run decision-making of persons within a household framework (Miller, 2005a, 2005b). This is an agent-based, microsimulation framework for modelling in a comprehensive and internally consistent manner both short-run (daily/weekly) activity/travel scheduling decisions and longer-run "resource management" decisions (change vehicle holdings, change residential location, change employment status/location, etc.) that form the core of land use modelling systems. Both persons and households are viewed as decision-making units (DMUs), where a DMU is defined as an intelligent agent that is capable of: perceiving the world around it; acquiring and using resources; scheduling its activities; and acting into the world. Figure 1 defines the key elements of a DMU, which consist of a perceptual interface that communicates with collaborating agents (e.g., person-household interactions) and the world in general; a knowledge base in which perceived information, agent tastes and preferences, agent memory, etc. are stored; three sub-classes of decision management agents (DMA), where a DMA is a specialized agent capable of making specific decisions. Specific DMAs within a DMU consist of: a set of project DMAs, in which each project consists of a logically connected set of activities directed towards a common goal (Axhausen, 1998); an activity scheduler that constructs provisional schedules for the DMU to execute, given activity episodes generated by the projects; and a stress/resource manager that allocates available resources (money, cars, houses, etc.) to projects and modifies these resources over time in response to stresses experienced by the DMU (Salvini and Miller, 2005; Miller, 2005b).