Distribution-independent inventory approach under multiple service level targets
A system and method are disclosed for an inventory planner that generates an inventory policy using any form of demand distributions, non-linear cost functions and/or multiple target measures of service levels, while taking into account a supply order lead time, such as, for example, a static or stochastic supply order lead time. The inventory policy generated by the inventory planner comprises an optimal and reproducible solution to one or more supply chain planning problems.
1. A system, comprising:
a computer comprising a processor and a memory, the computer configured to:
generate one or more transitional matrices based on received demand data, wherein the one or more generated transitional matrices comprise a possible number of items in inventory at each time step of one or more time steps;
generate one or more reward matrices, wherein the one or more reward matrices comprise one or more cost functions associated with each of one or more states;
model a supply chain planning problem over the one or more time steps as a Markov decision process based, at least in part, on the generated one or more transitional matrices and the generated one or more reward matrices;
solve the Markov decision process, by:
determining, from solution values, a resulting state after an order action is applied; and
generating from the solution values, indices comprising a minimal state and a target state;
generate an inventory policy based, at least in part, on a solution of the modeled supply chain planning problem, wherein the generated inventory policy comprises one or more order quantity decisions for each of the one or more time steps; and
in response to and based, at least in part, on the inventory policy, cause the one or more supply chain entities to ship one or more items.
2. The system of claim 1 , wherein the Markov decision process further comprises absorbing states, wherein a penalty is associated with the absorbing states to ensure the solution has no advantage to end in the absorbing states.
3. The system of claim 1 , wherein the received demand data comprises a demand probability distribution and wherein demand associated with the received demand data is time dependent.
4. The system of claim 1 , wherein at least one of the one or more generated transitional matrices incorporates a stochastic lead time.
5. The system of claim 1 , wherein the generated inventory policy is stationary with respect to the one or more time steps.
6. The system of claim 1 , wherein the generated inventory policy further comprises one or more order rules.
7. The system of claim 1 , wherein the computer is further configured to:
generate a demand probability as a weighted sum of probabilities of one or more lead time durations.
8. A method, comprising:
generating, by a computer comprising a processor and a memory, one or more transitional matrices based on received demand data, wherein the one or more generated transitional matrices comprise a possible number of items in inventory at each time step of one or more time steps;
generating, by the computer, one or more reward matrices, wherein the one or more reward matrices comprise one or more cost functions associated with each of one or more states;
modeling, by the computer, a supply chain planning problem over the one or more time steps as a Markov decision process based, least in part, on the generated one or more transitional matrices and the generated one or more reward matrices;
solving the Markov decision process, by:
determining, by the computer, from solution values, a resulting state after an order action is applied; and
generating, by the computer, from the solution values, indices comprising a minimal state and a target state;
generating, by the computer, an inventory policy based at least in part, on a solution of the modeled supply chain planning problem, wherein the generated inventory policy comprises one or more order quantity decisions for each of the one or more time steps; and
in response to and based, at least in part, on the inventory policy, cause, by the computer, one or more items to be shipped from an inventory.
9. The method of claim 8 , wherein the Markov decision process further comprises absorbing states, wherein a penalty is associated with the absorbing states to ensure the solution has no advantage to end in the absorbing states.
10. The method of claim 8 , wherein the received demand data comprises a demand probability distribution and wherein demand associated with the received demand data is time dependent.
11. The method of claim 8 , wherein at least one of the one or more generated transitional matrices incorporates a stochastic lead time.
12. The method of claim 8 , wherein the generated inventory policy is stationary with respect to the one or more time steps.
13. The method of claim 8 , wherein the generated inventory policy further comprises one or more order rules.
14. The method of claim 8 , further comprising:
generating, with the computer, a demand probability as a weighted sum of probabilities of one or more lead time durations.
15. A non-transitory computer-readable medium embodied with software, the software when executed is configured to:
generate one or more transitional matrices based on received demand data, wherein the one or more generated transitional matrices comprise a possible number of items in inventory at each time step of one or more time steps;
generate one or more reward matrices, wherein the one or more reward matrices comprise one or more cost functions associated with each of one or more states;
model a supply chain planning problem over the one or more time steps as a Markov decision process based, at least in part, on the generated one or more transitional matrices and the generated one or more reward matrices;
solve the Markov decision process, by:
determining, from solution values, a resulting state after an order action is applied; and
generating from the solution values, indices comprising a minimal state and a target state;
generate an inventory policy based at least in part, on a solution of the modeled supply chain planning problem, wherein the generated inventory policy comprises one or more order quantity decisions for each of the one or more time steps; and
in response to and based, at least in part, on the inventory policy, cause one or more items to be shipped from an inventory.
16. The non-transitory computer-readable medium of claim 15 , wherein the Markov decision process further comprises absorbing states, wherein a penalty is associated with the absorbing states to ensure the solution has no advantage to end in the absorbing states.
17. The non-transitory computer-readable medium of claim 15 , wherein the received demand data comprises a demand probability distribution and wherein demand associated with the received demand data is time dependent.
18. The non-transitory computer-readable medium of claim 15 , wherein at least one of the one or more generated transitional matrices incorporates a stochastic lead time.
19. The non-transitory computer-readable medium of claim 15 , wherein the generated inventory policy is stationary with respect to the one or more time steps.
20. The non-transitory computer-readable medium of claim 15 , wherein the generated inventory policy further comprises one or more order rules.