SARTS: A Dependable Real-Time Search Algorithm
ABSTRACT
Many imortant applications of AI systems need dependable real-time search
algorithms which are capable of archieving high deadline compliance and
high predictability. Deadline compliance represents the probability that
the system will meet a task's time constraints. Predictability represents
the system's ability to decide the feasibility of meeting the constraints
of a given task from a task set well ahead of the deadline. There is a
great need for evaluation and development of real-time search
algorithms. We propose a new algorithms, SARTS, that is based oon a novel
online technique to choose the proper values of parameters which control
the time allocated to planning based on the time constraints. SARTS also
provides criteria to predict its ability to meet the time constraints of
a given task. We evaluate the deadline compliance and predictability of
the proposed algorithm. The results of our experiments show that new
ideas incorporated in SARTS lead to higher performance in terms of
deadline compliance and predicatablity.
Keywords: Real-Time, On-Line Optimization,
Planning Effort Allocation, Search.
Please select the format you would like to view the paper
If your browser do not support the viewing of postscript file, it
will save as a file instead
![[HOME PAGE]](/research/shashi-group/home_btn.gif)
Design and Maintain by Julian Chow
Questions and Comments?
![[mailto:]](/research/shashi-group/icons/icnmail2.gif)
jchow@cs.umn.edu
[last Updated] Mon. July 24th 1995