1. Planning of Instance-based state Representation for Network Repairby
Aggelos Aggelidakis
Aggelos Aggelidakis Department ECE– Technical University of Crete

2. CSFR Model(Cost-Sensitive Fault Remediation)
Failure Mode
Take action
Gain information
Repair actions
Detect when correct functioning has been restored
Aggelos Aggelidakis Department ECE– Technical University of Crete

3. Aggelos Aggelidakis Department ECE– Technical University of Crete
CSFR Definition
S, fault states
Pr, prior probability of fault states
AT , test actions
AR, repair actions
C(s,a), cost functions
O(s,a), observation model
Aggelos Aggelidakis Department ECE– Technical University of Crete

4. Finding Optimal Policy
We can find an optimal repair policy via dynamic programming. Let B be the power set of S, which is the set of belief states of the system. For each b ∈ B and a ∈ AT ∪ AR, define the expected value of action a in belief state b as the expected cost of the action plus the value of the resulting belief state
Aggelos Aggelidakis Department ECE– Technical University of Crete

5. Value of resulting belief state
Where bi = {s ∈ b|o(s, a) = i} is the belief state resulting from taking action a in belief state band obtaining outcome i
V (b) = min a∈At∪Ar Q(b, a)
Pr(b)=∑ Pr(s), s∈B
c(b)=∑ Pr(s)*c(s,a)), s∈B, a∈At∪Ar
These equations can be evaluated in time proportional to
2|S| (|AT | + |AR |)(|S| + |AT | + |AR |), at which point the
optimal policy can be extracted by choosing the action a in
belief state b that minimizes Q(b, a). Such a policy will opti-
mally trade off actions to gain information, actions to repair,
and the information gained from a failed attempt at repair, in
a way that minimizes the total expected cost of repair
Aggelos Aggelidakis Department ECE– Technical University of Crete