University of Tsukuba
How tasks should be allocated to the human and the computer in the supervisory control is one of the central issues in research on human-machine systems. Human-centered automation ( Billings, 1991; Woods, 1989) assumes that human must be at the locus of control, and that the computer should be subordinate to the human. Does this mean, among various levels of autonomy listed in Table 1, the autonomy with levels 6 or higher must be denied? Do we have to assume that the level of autonomy must stay within the range of I through 5 at all times in every occasion?
|1.||The computer offers no assistance, human must do it all.|
|2.||The computer offers a complete set of action alternatives, and|
|3.||narrows the selection down to a few, or|
|4.||suggests one, and|
|5.||executes that suggestion if the human approves, or|
|6.||allows the human a restricted time to veto before automatic execution, or|
|7.||executes automatically, then necessarily informs human, or|
|8.||informs him after execution only if he asks, or|
|9.||informs him after execution if it, the computer, decides to.|
|10.||The computer decides everything and acts autonomously, ignoring the human.|
By using mathematical models, Inagaki ( 1991, 1993, 1995) has proposed the concept of situation- adaptive autonomy in which (i) the autonomy changes its level dynamically depending on the situation, and (ii) the autonomy with levels 6 or higher may be adopted for attaining system safety. It is proven that conditions exist for which the computer should be allowed to perform safety-control actions even when the human has given the computer no explicit directive to do so.
This paper discusses how authority can be traded between human and automation in a situation- adaptive manner. The following two problems are taken as examples: (i) Rejected takeoff problem in aviation, in which a human pilot must decide, upon an engine failure during the takeoff run, whether to continue the climb out or to abort the takeoff, and must take necessary control actions within a few seconds. (ii) Process control problem, where a human operator is requested to perform specified control tasks, and to take an appropriate action against a malfunction in the process.
Suppose an engine failure warning is set off while an aircraft is making its takeoff run. The pilot must decide whether to continue the takeoff (GO) or to abort it (NO-GO). The GO/NO-GO decision is made basically depending on whether the airspeed is less than a predesignated speed, V1. The standard procedure for the GO/NO-GO decision, upon an engine failure, is specified as follows: