Dynamic Allocation of Function

So far, the allocation of function between humans and machines could be assumed as being a one-shot effort allocation of function, once defined, remains static. Dynamic allocation of function is an alternative that makes it possible to adjust the division of labor between the human and the automation over time (Lee, 2006, p. 1581). This approach relies on two distinct antomation philosophies  

Where facilities exist in the automation to be able to undertake either type of automation, the guidance provided in BS/EN/ISO 11064-1 2001, combined with the human factors best practice to follow, will be helpful. 

It is important to note that attractive, dynamic allocation of function is not a panacea. Improper design can quite often magnify the detrimental effects, of which there 

---

Designers should recognize that operators are not inherently flawed and subsequently should not simply remove or automate as many tasks as possible. This approach is likely to lead to mental underload and could leave operators with incoherent sets of tasks. The goal instead should be to design the system and procedures to exploit the skills of human operators. Physical workload should be kept to a minimum, whereas mental workload should be optimized. Ways to optimize mental workload include the use of adaptive interfaces (systems that can detect periods of elevated workload and then assist the operator accordingly) and dynamic allocation of function (systems that relieve the operator of task elements during high workload portions of the task). 

A set of functions to be performed by humans and machines in different situations or under varying criteria (in the case of dynamic allocation of function)... 

BS/EN/ISO 11064-1 2001 has a considerable amount of human factors best practice already embedded within it. It is, of course, possible to benefit still further, particularly in terms of the methods (and sequence of methods) that can be used to undertake functional analyses, task analyses, actual allocation of function, and the diagnosis of cognitive and affective criteria. The standard touches on the concept of dynamic allocation of function and this is explored further in this section. 

This type of analysis provides further insight for allocation of function decisions. Activities A and C might indicate situations where automation has to occur (and where it typically does occur). This is a useful check in terms of mandatory assignments. Activity B, on the other hand, shows a similar pattern for situation 1 (perhaps automation has to, and typically does, occur here) but not for situations 2 and 3. Activity could occur in these situations (but typically does not), yet the allocation of function decisions might be quite different in these cases. This, then, is one route to systematically considering dynamic allocation of function—that is, a form of automation that is situation dependent. 

Cognitive Woik Analysis also helps to facilitate dynamic allocation of function. This feature can be demonstrated with a contextual activity template derived from the same scenario. Figure 7.12 plots the anticipated modes of system operation specified in... 

In summary, the HTA tool is presented as a means by which a significant evaluation of basic requirements and preliminary allocations can be undertaken. The two CWA-based approaches are presented as means to extend that analysis beyond basic requirements and increase the quality of technical critique—in particular, positioning the analysis so dynamic allocation of function can be facilitated. 

This form of implanned manual operation is unsatisfactory on a number of counts. The fact that the operator may normally be insulated from the process by the automatic control systems means that he or she will probably not be able to develop the knowledge of process dynamics ("process feel") necessary to control the system manually, particularly in extreme conditions. Also, the fact that manual control was not "designed into" the systems at the outset may mean that the display of process information and the facilities for direct control are inadequate. A number of techniques are available to assist designers in the allocation of function process. Some of these are described in Meister (1985). In a paper entitled "Ironies of Automation" Bainbridge (1987) notes four areas where the changed role of the human in relation to an automated system can lead to potential problems. These will be discussed below. 

In the case of dynamic allocation, there is an additional requirement to consider how these heuristics apply to the same functions but in different situations for example, a task could conceivably be repetitive and monotonous in one situation but encourage a feeling of usefulness in another. 

The process view (behavior or functional view) describes the dynamic aspect of systems as well as the behavior of elements at their intersections to each other and in a defined environment. Relations could be any kind of communication (technical but also man-machine communication etc.) time behavior as well as allocation and structure aspects such as parallelism, distribution, integration, performance and scalability. Activity, sequence or timing diagrams can be used as UML-diagrams. 

The method described in this article has been converted into computer code, as the program MOZYME. MOZYME is written in FORTRAN-77, and uses dynamic memory allocation. As far as possible, the names of routines in MOZYME are similar to those in the equivalent program that uses conventional methods, MOPAC. At present, because the method is still very new, the range of functions of MOZYME is very limited. However, enough functions exist to allow the validity of the new method to be tested. 

GOOS is a comprehensive, full-scale, environmental monitoring system aimed at solving a wide spectrum of problems (Holland and Nowlin, 2001 GOOS, 2002). The history and prospects for GOOS development are reflected in the dynamics of its structure and functions. For example, in 2000 deployment of its equipment constituted only 30% of the level planned for the following decade, but by 2009 all its planned levels look likely to be allocated and equipped with the necessary instruments to measure a wide range of characteristics of the World Ocean and the atmosphere. 

An additional problem with free allocation is that it can lead to rather perverse dynamic incentives. For instance, if future allowances are allocated as a function of present emission levels, firms... 

The resource management and real-time control functions of FMS are closely related to the dynamic scheduling system. The resource-management system should be activated by a dynamic scheduling system to allocate resources to production process to achieve real-time control for FMS. The resources to be controlled involve tools, automatic guided vehicles, pallets and fixtures, NC files, and human resources. 

In this implementation, it is crucial to be able to dynamically manage memory allocation. In fact, each subinterval can be further split and, every time this is done, a space has to be created to collect the function values in the six new points required by the procediu-e. When the error is good on a portion of the interval, all the memory allocated for that portion is no longer necessary and can be promptly freed up and made available. 

The working environments of task allocation can be static or dynamic [4]. Static task allocation assumes completely known information about the environment, such as the number of tasks and robots, the arrival time of tasks, and the process of task execution. Traditionally, applications in multi-robot domains have largely remained in static scenarios, with an aim to minimise a cost function, such as total path length, or execution time of the team. Obviously, static approaches cannot adapt to changes in a dynamic environment. 

Despite the large amount of work and despite standardization efforts, the use of HDLs for synthesis is still an unresolved issue. Although VHDL [36] has emerged as a standard for hardware specification, some questions remain as to its applicability as an input language for high-level synthesis. For example, it is unclear what class of applications should be described in a sequential or a functional style, for behavioral specification and synthesis. Furthermore, several language constructs, well suited for simulation, are difficult to synthesize (e.g., dynamic memoiy allocation and sensitivity lists [18]), or are not applicable for synthesis (e.g. assertions). 

There are some safe subsets of Ada, such as Spark Ada, which are more restricted for special safety applications and use on safety-critical systems. These subsets restrict some of the standard language functions and syntax that could be somewhat risky because they are difficult to verify, such as multithread processing, dynamic memory allocation, and recursion. 

---