Decentralized Time Geography for Ad-Hoc Collaborative Planning

For an autonomous physical agent, such as a moving robot or a person with their mobile device, performing a task in a spatio-temporal environment often requires interaction with other agents. Such interaction can be initiated by ad-hoc collaborative planning and decision-making, which then leads to physical support on site. Ad-hoc communication between agents, if performed in a peer-to-peer manner, does not involve a central instance and facilitates some form of local knowledge in each agent by nature. In this paper we introduce the notion of decentralized time geography, which differs from the traditional time-geographic framework by taking into account limited local knowledge. This allows us to perform a space-time analysis within a timegeographic framework that represents local knowledge in a distributed environment as required for ad-hoc coordinated action between agents in physical space. More specifically, we investigate the impact of general agent movement, replacement seeking, and location and goal-directed behavior of the initiating agent on the outcome of the task. Empirical tests in a multi-agent simulation framework provide both a proof of concept and specific results for different combinations of agent density and communication radius.