Analysing Dynamic Function Scheduling Decisions

Function allocation, as a process used in the construction of dependable complex systems, is growing in significance. It involves a documented and rational process for deciding what aspects of the system should be controlled by which human roles in the system and how the system should be automated to support these roles more effectively. As computer systems have become more advanced and the control of systems more complex dynamic function allocation has become increasingly desirable in situations where the automation may take over or give back function to the human user. This position statement briefly discusses an approach to dynamic function allocation called dynamic function scheduling where dynamic scheduling of activities takes place on the time dimension and discusses the role that tools may play in helping identify dependable dynamic function scheduling solutions. Complex work systems typically involve teams of people co-operating and using technology to achieve work goals. These goals are usually achieved under time constraint. In order to achieve them in a timely and reliable manner, the implementation of the functions that achieve the goals may vary according to situation. This topic is dealt with in research into dynamic function allocation (see [Hancock and Scallen, 1998] and [Scerbo, 1996] for an overview). The overall focus of this work is how automation can be used adaptively, according to the current demands on the system, and the capabilities and workload levels of the agents involved, in order to offer optimal support to the human operator. The problem of function allocation is to take a set of functions that describe the work that the system is to do, in the contexts in which the work is to be carried out, and to decide how these functions should be implemented by the roles that are defined within the system. Methods are required that will enable system engineers to take work descriptions and to consider how the actions within them should be implemented. Typically the methods that exist are concerned with static allocations in the sense that the decision about how roles are allocated to function occur at design time. In practice, it makes sense to consider the appropriateness of different configurations in different situations under different conditions of workload and different requirements of situation awareness for example. Hence an in-car navigation system may have a different level of automation in which certain default inputs are presumed when the car is moving or in gear than when the car is stationery and in neutral. In addition to sharing and trading functions among humans and automation, it may be possible to change the way functions are allocated in time in order to meet the required deadline. Of particular relevance for designing function scheduling processes is a better understanding of temporal awareness ([Grosjean and Terrier, 1999]) and of the use of time as information ([Michon, 1990, Block, 1990]). Analysis techniques are required to aid the process of choosing particular dynamic function scheduling designs [Hildebrandt and Harrison, 2002] where: