| Stankovic, J., K. Ramamritham, and D. Niehaus, "On Using the Spring Kernel to Support Real-Time AI Applications," Proc. Euromicro Workshop on Real-Time Systems, June 1989. |
....has been very simple. It is usually assumed that the application passes tasks on to the scheduler for scheduling and does not react when the scheduler is unable to schedule some tasks before their deadlines or only able to provide low quality solutions. A more complex interface is proposed in [Stankovic et al. 1989] that allows the application to ask what if questions of the scheduler and modulate the behavior of the scheduler, however these ideas have not been implemented to date. In our work on a complex, real time, multi agent problem solver we have found that a more bidirectional, negotiation based ....
....approach is described in [Hudlick a and Lesser, 1987] where a diagnosis system watches and informs the scheduling process. Another layered architecture (briefly mentioned above) that has been proposed for real time AI tasks is the extensions to the Spring real time operating system described in [Stankovic et al. 1989] . In this work, elements of the interface between the AI level and the scheduler include time planners that allow the AI level to ask what if scheduling questions without disrupting the low level scheduling. The work on the Spring operating system also supports guarantees and endorsements, ....
J. A. Stankovic, K. Ramamritham, and D. Niehaus. On using the Spring kernel to support real-time AI applications. In Proceedings of the EuroMicro Workshop on Real-time Systems, 1989.
....environment. More recent design to time research has focused on heuristic scheduling techniques and on the interface between a design to time scheduler and the application that it is embedded in. For another perspective on the problem of real time scheduling of AI tasks see Stankovic et al. [Stankovic et al. 1989] . 3.3 Time constrained reasoning under uncertainty ###### ######## ## ## ## ## ####### ##### ###### ##### #### ###### ####### ####### ## ## ## ## ## ## ## ## ## ## ## ## ## ## ## ## ## ## ## ## ## ## ## ## ###### #### ## ## # # ## # # # # # # # # # # # # # # ## # ## ## ##### ##### # Figure ....
J. A. Stankovic, K. Ramamritham, and D. Niehaus. On using the Spring kernel to support real-time AI applications. In Proceedings of the EuroMicro Workshop on Real-time Systems, 1989.
....has been very simple. It is usually assumed that the application passes tasks on to the scheduler for scheduling and does not react when the scheduler is unable to schedule some tasks before their deadlines or only able to provide low quality solutions. A more complex interface is proposed in [Stankovic et al. 1989] that allows the application to ask what if questions of the scheduler and modulate the behavior of the scheduler, however these ideas have not been implemented to date. In our work on a complex, real time, multi agent problem solver we have found that a more bidirectional, negotiation based ....
....addition of hard deadlines and more flexible temporal commitments. However, the algorithm is only real time in the same sense as the underlying local scheduler (developed by Garvey [Garvey and Lesser, 1994] and not in the strong sense of guaranteed computation like the Spring scheduler 146 [Stankovic et al. 1989] (how these all can work together is a future research direction, see also the discussion in Section 5.7) Because GPGP is a cooperative, team oriented mechanism, it assumes that no agent has a consistently better view of problem solving (either because of the problem solving algorithm or by some ....
J. A. Stankovic, K. Ramamritham, and D. Niehaus. On using the Spring kernel to support real-time AI applications. In Proceedings of the EuroMicro Workshop on Real-time Systems, 1989.
....statistics, keeping information for debugging or on line decision making, performing computation to decide what computation to pursue next (or for the next interval) self optimization, and self modification. Features covering all these examples appear or are planned for the Spring Kernel [14, 6]. Much of the reflective capabilities of the Kernel arise from the task management and scheduling features of the Kernel. We restrict our discussion to these areas. 4.1 Reflection in Task Management and Scheduling Tasks arise when real time programs specified in the form of communicating ....
....degraded service available from the system, or notifying the system managers that additional processing power must be added to the system. To these basic scheduling modules we plan to add one or more Time Planners which have access to the task descriptors, and a condition monitor facility [14]. The Time Planners could be considered versions of the local scheduler being invoked for different purposes, i.e. to suggest tradeoffs, to perform planning of future schedules for tasks, or to suggest causes of the problems in the current plan. Multiple Time Planners may be invoked ....
Stankovic, J., K. Ramamritham, and D. Niehaus, "On Using the Spring Kernel to Support Real-Time AI Applications," Proc. Euromicro Workshop on Real-Time Systems, June 1989.
.... community, ffl what languages and language features are appropriate, ffl what RTAI and system level paradigms [14] are or are not appropriate and how do they affect each other, ffl what interfaces should exist between the OS and the reactive front end and between the OS and the cognitive layer [13], ffl how should various subsystems be integrated such as vision and robot control, ffl can we solve the dilemma of wanting a predictable, analyzable, safely behaving system, but one which operates in highly uncertain environments and learns, ffl how to exploit high performance parallel ....
J. Stankovic, K. Ramamritham and D. Niehaus, On Using The Spring Kernel To Support Real-Time AI Applications, Proc. EuroMicro Workshop on Real-Time Systems, June 1989.
No context found.
Stankovic, J. A., K. Ramamritham, and D. Niehaus, "On Using the Spring Kernel to Support Real-Time AI Applications," Proceedings Euromicro Workshop on Real Time, pp. 51-60, Como, Italy, June 1989.
Online articles have much greater impact More about CiteSeer.IST Add search form to your site Submit documents Feedback
CiteSeer.IST - Copyright Penn State and NEC