Dealing with the challenges of distributed planning in a stochastic environment: coordinated contingency planning

Because of its cognitive complexity, the responsibility for operating the National Airspace System (NAS) is distributed among many organizations and individuals. An understanding of how this distributed work system functions requires consideration not only of the allocation of control and responsibility, but also of the distribution of data, knowledge, processing capacities and characteristics, goals and priorities. It further requires consideration of how alternative architectures for distributing work (as defined by these different dimensions) impact performance on different types of tasks described by P. J. Smith et al., (1999). Given such a distributed system, one of the most significant challenges is how to plan at a system level in the face of uncertainty, where the level of uncertainty changes over time. In this paper, we explore a future vision that allows traffic managers and dispatchers to more fully communicate their beliefs about possible weather and traffic constraints described by P. J. Smith et al., (2003), and to indicate how they would respond (or would prefer as a response by someone else who has control and responsibility for responding at that point), depending upon the state of their knowledge at the time when they must act. Instead of communicating their beliefs about the single most likely scenario (weather at BYP from 1400-1500Z), they would communicate their beliefs about the range of possible scenarios that could potentially occur. Then, instead of communicating a single plan, the traffic managers and dispatchers would communicate preferences, constraints or intentions (depending upon whom would have control to make the relevant decision) for each of these contingencies. In addition, because the degree of uncertainty about weather and traffic varies over time and often becomes smaller as the