Continuous Inventory Policies with Non-Stationary and Stochastic

Demand

Abstract here!

Key words: Non-Stationary demand, time-dependent demand, Phase-type distribution, inventory

systems, ordering policies

Introduction

Model

: interval length.
N : Number of intervals of length .
The reoder point at time t is s(t) and the reorder level is S(t). At time t = n , the reorder policy
is (s(t), S(t)) = (sn , Sn) for n = 1, 2, , N . The time dependent reorder policy (s(t), S(t)) is
where s = {s1 , s2 , . . . , sm } and S = {S1 , S2 , . . . , Sm}.
represented by (
s, S)

System State Probabilities - Inventory Position

where s = {s1 , s2 , . . . , sm } and S = {S1 , S2, . . . , Sm },

For a given dynamic reorder policy, (
s, S)
the state of the dynamic inventory position at time t is represented by the value of the inventory
1

position at time t and the state of the P ht process at time t.

The inventory position is IP(t) {s(t) + 1, . . . , u}, where u = max(Sn ), and the state of the P ht
process is A(t) {1, 2, , m1 +m2 }. Let Pi,n (t) = P(IP(t) = i, A(t) = n) for i = s(t)+1, . . . , u and
n = 1, . . . , m1 + m2 . The resulting inventory position Kolmogorov Forward Equations (IP-KFEs)
are presented in Equations 1 to 6.
For 1 < n m1 and s(t) < i S(t),

(t) = m1 1 (t) Pi,n(t) + m1 1 (t) Pi1,n .
Pi,n

(1)

For m1 + 1 n (m1 + m2 ) and s(t) < i S(t),


(t) = m2 2 (t) Pi,n(t) + m2 2 (t) Pi1,n .
Pi,n

(2)

For n = 1 and s(t) < i < S(t),


(t) = m1 1 (t) Pi,1 (t) + (t) m1 1 (t) Pi1,m1 + (t) m2 2 (t) Pi1,m1 +m2 .
Pi,1

(3)

For n = 1 and i = S(t),


(t) = m1 1 (t) PS(t),1(t) + (t) m1 1 (t) Ps(t)+1,m1 + (t) m2 2 (t) Ps(t)+1,m1 +m2 .
PS(t),1

(4)

For n = m1 + 1 and s(t) < i < S(t),


(t) = m2 2 (t) Pi,m1 +1 (t) + (1 (t)) m1 1 (t) Pi1,m1
Pi,m
1 +1
+(1 (t)) m2 2 (t) Pi1,m1+m2 .

(5)

For n = m1 + 1 and i = S(t),


(t) = m2 2 (t) PS(t),m1+1 (t) + (1 (t)) m1 1 (t) Ps(t)+1,m1
PS(t),m
1 +1
+(1 (t)) m2 2 (t) Ps(t)+1,m1 +m2 .

References

(6)