A distributed planning system for air traffic control
Ron Lo · 1988
A Distributed Planning System (DPS) is a network whose nodes represent distinct processors, each cooperating with a selected set of others to achieve a common set of goals. In such a system, the nodes are decentralized, loosely-coupled, and have varying degrees of autonomy. Nodes are grouped (partitioned) into sub-networks (groups) dynamically according to the characteristics of new problems (tasks). Nodes within a group are organized according to a specific structure. The research reported here investigates the effectiveness of using different organizational structures in DPS for Distributed Air Traffic Control (DATC). Distributed organizational and control techniques are used to satisfy the major objective--all aircraft get to their destination safely. However, conflicts can happen; for example two or more aircraft may get too close to each other. A resolution-plan needs to be devised to resolve conflicts. A group of aircraft is identified with reference to their location, and organized according to a specific structure to resolve the detected potential conflicts. The Coordinator-Coworker structure is used as a generic organization. The nodes in it follow a dynamic pattern of self-organization in terms of the roles (coordinator, coworker and nominator) they play. A coordinator-coworker hierarchy of level n means there are at most $n$ levels of coordinator-coworker relationship. The DATC testbed, a simulation model for multiple aircraft, has been implemented and used to explore various methodologies empirically. Organizational structures with coordinator-coworker hierarchy of level one and two have been simulated and compared with the local centralized approach. (The latter can be viewed as coordinator-coworker hierarchy of level zero.) The results of these studies fall into three categories: (i) mechanisms aimed at the connection, communication, uncertainty and coherence problems; (ii) building a general-purpose testbed for Distributed Artificial Intelligence (DAI) research; and (iii) investigating the feasibility of delegating planning responsibilities to individual aircraft for DATC.