WO2017151087A1

WO2017151087A1 - Select recovery plan of subset

Info

Publication number: WO2017151087A1
Application number: PCT/US2016/020019
Authority: WO
Inventors: Omer Barkol; Tomer SAGI; Inbal Tadeski; Guy WEINER
Original assignee: Hewlett Packard Enterprise Development Lp
Priority date: 2016-02-29
Filing date: 2016-02-29
Publication date: 2017-09-08

Abstract

A set of recovery plans may be calculated and stored based on a set of possible disruptions to a base plan. Solutions to the set of recovery plans may be incrementally improved during an offline time period. Each of the set of recovery plans may be calculated for a combination of the set of possible disruptions. A subset of the stored set of recovery plans may be selected during an online time period after the offline time period, based on a match between a set of actual disruptions and the set of possible disruptions. At least one of the subset may continue to be incrementally improved during the online time period, if at least one of the subset has a cost that is reducible. One of the recovery plans of the subset may be selected based on a cost calculated for the subset of recovery plans.

Description

SELECT RECOVERY PLAN OF SUBSET

BACKGROUND

[0001 ] Enterprises often run complex operations requiring elaborate operations-management methods. Operations Research (OR) literature includes numerous algorithms designed to optimize operations management.

BREF DESCR!PTSON OF THE DRAWINGS

[0002] The following detailed description references the drawings, wherein:

[0003] FIG. 1 is an example block diagram of a device to select a recovery plan of a subset;

[0004] FIG. 2 is another example block diagram of a device to select a recovery plan of a subset;

[0005] FIG. 3 is an example block diagram of a computing device including instructions for selecting a recovery plan of a subset; and

[0006] FIG. 4 is an example flowchart of a method for selecting a recovery plan of a subset.

DETAILED DESCRIPTION

[0007] Specific details are given in the following description to provide a thorough understanding of embodiments. However, it will be understood that embodiments may be practiced without these specific details. For example, systems may be shown in block diagrams in order not to obscure embodiments in unnecessary detail. In other instances, well-known processes, structures and techniques may be shown without unnecessary detail in order to avoid obscuring embodiments.

[0008] Enterprises often use automated tools to manage and optimize their operations. Many of these approaches are based upon Artificial intelligence (Al) techniques. However, the optimal plan is rarely carried out as-is. Real-world disruptions force the enterprise to adapt the plan to the changing situation at real time, often leading to sub-optimal results. Unfortunately, any substantial operation is bound to be affected by disruptions. For example, airlines optimize their flights schedule and aircraft rotations for maximal revenue, but frequently change these plans in face of malfunctions, delays, and weather conditions.

[0009] Further, a commercial flight-schedule may easily go awry when an aircraft breaks down or bad weather causes airport closure. Thus, preparing a single, albeit optimal, plan does not constitute a robust approach to running a complex operation. Ideally, automated tools should provide the enterprise with a range of alternatives, allowing operations to be carried out within predictable limits to plan quality degradation and increased cost with respect to the original (i.e. optimal) plan. Furthermore, achieving such robustness by wrapping an existing solution rather than re-writing it, is preferred, as some algorithms are highly specialized to their domain, and cannot be easily changed.

[0010] To address this vulnerability, robust Ai approaches to planning and scheduling have been sporadically suggested. However, few works offer meaningful computational support for recovering from schedule breakdowns. In many industries, even though disruptions cause major financial implications for companies, recovery is done via manual, ad-hoc decisions taken by local operators. However, while some academic effort has been directed at suggesting schedule recovery mechanisms, the solutions suggested are often partial, limited to small-scaled problems and have been slow on the uptake by industry.

[001 1 ] Examples provide a generic method for adding robustness to operations management algorithms by dividing the re-planning task into two phases: an offline phase and an online phase, in the offline phase, where the actual set of disruptions is unknown but time is less of a factor, a repository of partial plans is constructed. In the online phase, some of these plans are fetched in order to assist in the short time of online re-planning.

[0012] In one example, a set of recovery plans may be calculated and stored based on a set of possible disruptions to a base plan. Solutions to the set of recovery plans may be created and incrementally improved upon during an offline time period. Each of the set of recovery plans may be calculated for a combination of the set of possible disruptions. A subset of the stored set of recovery plans may be selected during an online time period after the offline time period, based on a match between a set of actual disruptions and the set of possible disruptions. At least one of the subset may continue to be incrementally improved during the online time period, if at least one of the subset has a cost that is reducible. One of the recovery plans of the subset may be selected based on a cost calculated for the subset of recovery plans.

[0013] Thus, examples provide improved solutions within a similar time frame to online only versions, by preparing a repository of partial solutions in advance during an offline phase. Since the online phase of examples starts their search from pre- processed starting points, the solutions are generally of higher quality, e.g. incur lower penalties.

[0014] Referring now to the drawings, FIG. 1 is an example block diagram of a device 100 to select a recovery plan of a subset. The device 100 may include or be part of a microprocessor, a controller, a memory module or device, a notebook computer, a desktop computer, an all-in-one system, a server, a network device, a wireless device, a network and the like.

[0015] The device 100 is shown to include an online unit 1 10 and an offline unit 120. The online and offline units 1 10 and 120 may include, for example, a hardware device including electronic circuitry for implementing the functionality described below, such as control logic and/or memory, in addition or as an alternative, the online and offline units 1 10 and 120 may be implemented as a series of instructions encoded on a machine-readable storage medium and executable by a processor.

[0016] The offline unit 1 10 may calculate and store a set of recovery plans 212 based on a set of possible disruptions 1 14 to a base plan 1 18 during an offline time period 1 18. Further, the offline unit 1 10 may incrementally improve solutions to the set of recovery plans 1 12 during the offline time period 1 18. Each of the set of recovery plans 1 12 may be calculated for a combination of the set of possible disruptions 1 14.

[0017] The online unit may select a subset 122 of the stored set of recovery plans 1 12 during an online time period 128 after the offline time period 1 18, based on a match between a set of actual disruptions 124 and the set of possible disruptions 1 14. The online unit 120 may continue to incrementally improve at least one of the subset 122 during the online time period 128, if at least one of the subset 122 has a cost 126 that is reducible. The online unit 120 may select one of the recovery plans of the subset 122 based on a cost 126 calculated for the subset of recovery plans 122. The device 100 is explained in greater detail below with respect to FIGS. 2-4.

[0018] FIG, 2 is another example block diagram of a device 200 to select a recovery plan of a subset. The device 200 may include or be part of a microprocessor, a controller, a memory module or device, a notebook computer, a desktop computer, an all-in-one system, a server, a network device, a wireless device, a network and the like. Further, the device 200 of FIG. 2 may include at least the functionality and/or hardware of the device 100 of FIG. 1 .

[0019] For example, an offline unit 210 and an online unit 220 of the device 200 of FIG. 2 may include at least the respective functionality and/or hardware of the offline and online units 1 10 and 120 of the device 100 of FIG. 1 . The device 200 of FIG. 2 is also shown to interface with a repository 230. The repository 230 may include, for example, random-access memory device (RAM), non-volatile random- access memory (NVRAM) such as memrisfor technology, a flash drive, a disk drive, a network-connected storage element and the like.

[0020] As noted above, the offline unit 210 may calculate and store a set of recovery plans 212 based on a set of possible disruptions 1 14 to a base plan 1 16 during an offline time period 1 18. The offline unit 210 may store the set of recovery plans 212 at the repository 230. Further, the offline unit 210 may incrementally improve solutions to the set of recovery plans 212 during the offline time period 1 18. Each of the set of recovery plans 212 may be calculated for a combination 214 of the set of possible disruptions 1 14. [0021 ] Each combination 214 of the set of possible disruptions 1 14 may have more than one solution and therefore be associated with multiple recovery plans. Thus, while each recovery plan may be associated with one combination 214 of the set of possible disruptions 1 14, each set of possible disruptions 1 14 may be associated with one or more recovery plans.

[0022] The offline unit 210 may select an order to calculate the set of recovery plans 212 based on which of the set of possible disruptions 1 14 has a greater likelihood to occur and/or takes a greater time to re-plan around. The offline unit 210 may incrementally improve the set of recovery plans 212 by iterativeiy re-planning the base plan 1 16 such that the cost 128 of the corresponding recovery plan is non- increasing as computation time increases.

[0023] The set of recovery plans 212 may include alternate plans to recover from a breakdown of the base plan 1 16 caused by the set of possible disruptions 1 14. Also, the set of recovery plans 212 are to meet constraints 220 of the base plan 1 16. The offline unit 210 may incrementally calculate the set of recovery plans 212 to have a lower cost 128.

[0024] The online unit 220 may select a subset 122 of the stored set of recovery plans 212 during an online time period 128 after the offline time period 1 18, based on a match between a set of actual disruptions 124 and the set of possible disruptions 1 14. For example, the set of recovery plans 212 may be stored at the repository 230 before being accessed by the online unit 220.

[0025] The online unit 220 may continue to incrementally improve at least one of the subset 122 during the online time period 128, if at least one of the subset 122 has a cost 126 that is reducible. The online unit 220 may select one of the recovery plans of the subset 122 based on a cost 126 calculated for the subset of recovery plans 122. For example, the online unit 220 may select the recovery plan having a lowest of the costs 126 of the subset 122. The online unit 220 may select the subset 122 when at least one disruption is added to the set of actual disruptions 124. The at least one disruption may be added to the set of actual disruptions 124 when the at least one disruption at least one of has occurred and will occur.

[0026] The offline and online time periods 1 18 and 128 occur at separate times. For example, the online time period 138 may occur days after the offline time period 1 18. The online time period 128 may be smaller than the offline time period 1 18. Further, the offline time period 1 18 may be a considerably longer time than the online time period 128. For example the offline time period 1 18 may span multiple hours and/or days while the online time period 128 may be 5-10 minutes.

[0027] The online unit 220 may select the subset 122 of the set of recovery plans 212 based on a greatest number of the disruptions that match between the sets of possible and actual disruptions 1 14 and 124. The online unit 220 may stop incrementally re-planning the cost 126 of at least one of the subset 122 if the online time period 128 expires.

[0028] The cost 126 of each of the set of recovery plans 212 may be based on a cost function 222 of the corresponding recovery plan. The cost function 222 may relate to a weighted 224 sum of consequences caused by the recovery plan in response to the corresponding combination 214 of the set of possible disruptions 1 14.

[0029] The consequences having a greater likelihood to occur and/or greater penalty for re-planning may be associated with a greater weight 224, Examples may also include other polices for dictating the weight 224, such as commercial considerations like reducing known high customer dissatisfaction. The offline unit 210 may incrementally reduce the cost calculated by cost function 222 of the set of possible recovery plans 1 12. The online unit 220 may incrementally further reduce the costs calculated by cost function 222 of the subset of recovery plans 122, if the costs of the subset of recovery plans 122 are further reducible. For example, the offline and/or online units 210 and 210 may seek to reduce the cost 126 via optimizing planning algorithms, such as A* or iterative deepening.

[0030] In one example, the base plan 1 18 may relate to a set of flights to be operated by an airline over a given time period. Here, the offline and/or online units 210 and 220 may incrementally reduce the cost 222 by incrementally revising the set of flights while meeting constraints 220 of the base plan 1 16 and overcoming any of the disruptions of at least one of the set of possible and actual disruptions. Examples of revising the set of flights may include delaying, cancelling or adding flights, rebooking passengers to different flights, changing assignments (e.g. a different plane or crew taking a given flight) and the like.

[0031 ] in this case, the constraints 220 may be associated with at least one of an aircraft, an airport, itineraries, turn-around times, and a crew. For example, the constraints 220 may relate to seating capacities, maintenance, airport capacities, minimum connection time, minimum turn-round time, transit time, and the like.

[0032] Further, the set of possible disruptions 1 14 may include a flight, aircraft, airport and/or crew disruption. The flight disruption may relate to a flight being delayed or cancelled. The aircraft disruption may relate to a time the aircraft is unavailable. The airport disruption may relate to temporarily reduced departure and arrival capacities of the airport. Also, the cost function 222 may relate operating costs, passenger inconvenience costs, and/or inconsistency costs that are incurred if the positions of the aircraft at the end of a recovery period do not match a planned position,

[0033] Thus, disruptions 1 14 do not have a cost 128 but the consequences of disruptions 1 14 do have a cost 126. For example, the same disruption (e.g a 30-min. delay in take-off) may result in a relatively cheap or expensive recovery plan 122, depending on the consequences of the delay. Therefore, the cost 128 may depend on the plan implied by the solution of the recovery plan 122 (e.g. the cost of re-routing planes, cancelling itineraries, ferrying empty aircrafts around, etc.) and not the disruptions 1 14 causing the re-planning. For instance, the same type of disruption 1 14 may result in recovery plans 1 13 with very different costs 126.

[0034] Re-planning algorithms are often iterative, capable of producing better solutions given more time. However, due to time constraints, they are given inadequate time to do so, resulting in sub-optimal plans. As the scale of the problem grows, the time required to reach the optimal solution grows rapidly and actual solutions given limited time become even more sub-optimal.

[0035] In examples, the off-line phase (e.g. the offline unit 210 during the offline time period 1 18) may result in a repository 230 of alternate recovery plans 212, designed to satisfy those possible disruptions 1 14 most likely to occur, and most time consuming for which to re-plan. In the on-line phase (e.g. the online unit 220 during the online time period 128), given existing or imminent actual disruptions 124, a new recovery plan 122, based upon the prepared repository 230 of alternate recovery plans 212 that is consistent with the actual disruptions 124 may be rapidly returned.

[0036] FIG. 3 is an example block diagram of a computing device 300 including instructions for selecting a recovery plan of a subset, in the embodiment of FIG. 3, the computing device 300 includes a processor 310 and a machine-readable storage medium 320. The machine-readable storage medium 320 further includes instructions 321-325 for selecting the recovery plan of the subset.

[0037] The computing device 300 may be included in or part of, for example, a microprocessor, a controller, a memory module or device, a notebook computer, a desktop computer, an all-in-one system, a server, a network device, a wireless device, or any other type of device capable of executing the instructions 321-325. in certain examples, the computing device 300 may include or be connected to additional components such as memories, controllers, etc.

[0038] The processor 310 may be, at least one central processing unit (CPU), at least one semiconductor-based microprocessor, at least one graphics processing unit (GPU), a microcontroller, special purpose logic hardware controlled by microcode or other hardware devices suitable for retrieval and execution of instructions stored in the machine-readable storage medium 320, or combinations thereof. The processor 310 may fetch, decode, and execute instructions 321 -325 to implement selecting the recovery plan of the subset. As an alternative or in addition to retrieving and executing instructions, the processor 310 may include at least one integrated circuit (IC), other control logic, other electronic circuits, or combinations thereof that include a number of electronic components for performing the functionality of instructions 321 -325. [0039] The machine-readable storage medium 320 may be any electronic, magnetic, optical, or other physical storage device that contains or stores executable instructions. Thus, the machine-readable storage medium 320 may be, for example, Random Access Memory (RAM), an Electrically Erasable Programmable Read-Only Memory (EEPROM), a storage drive, a Compact Disc Read Only Memory (CD-ROM), and the like. As such, the machine-readable storage medium 320 can be non-transitory. As described in detail below, machine- readable storage medium 320 may be encoded with a series of executable instructions for selecting the recovery plan of the subset.

[0040] Moreover, the instructions 321-325, when executed by a processor (e.g., via one processing element or multiple processing elements of the processor) can cause the processor to perform processes, such as, the process of FIG. 4. For example, the prepare instructions 321 may be executed by the processor 310 to prepare a set of recovery plans based on a base plan and a set of possible disruptions. The set of recovery plans may re-plan the base plan to overcome combinations of the set of possible disruptions.

[0041 ] The reduce cost set instructions 322 may be executed by the processor 310 to incrementally reduce a cost of the set of recovery plans during an offline time period. The select subset instructions 323 may be executed by the processor 310 to select a subset of the set of recovery plans having a closest match to a set of actual disruptions. The closest match may be based on a number and/or importance of any of the disruptions matching between the set of possible and actual disruptions. [0042] The reduce cost subset instructions 324 may be executed by the processor 310 to incrementally further reduce the cost of at least one of the subset of recovery plans during an online time period that is after the offline time period. The cost may be based on consequences of the recovery plan to overcome the corresponding combination of possible disruptions,

[0043] The cost may be incrementally reduced by iterativeiy swapping pairs of elements of the recovery plan while meeting a constraint of the base plan and continuing to overcome the corresponding combination of possible disruptions. The select recovery plan instructions 325 may be executed by the processor 310 to select the recovery plan of the subset of having a lowest of the costs.

[0044] FIG. 4 is an example flowchart of a method 400 for selecting a recovery plan of a subset. Although execution of the method 400 is described below with reference to the device 200, other suitable components for execution of the method 400 can be utilized, such as the device 100. Additionally, the components for executing the method 400 may be spread among multiple devices (e.g., a processing device in communication with input and output devices), in certain scenarios, multiple devices acting in coordination can be considered a single device to perform the method 400. The method 400 may be implemented in the form of executable instructions stored on a machine-readable storage medium, such as storage medium 320, and/or in the form of electronic circuitry. Further, the method 4000 may be computer-implemented, such as via distributed or cloud-based computing.

[0045] At block 410, the device 200 incrementally improves solutions for a set of recovery plans 212 based on a base plan 1 18 and a set of possible disruptions 1 14. Each of the set of recovery plans 212 may provide an alternate solution to overcoming a combination 214 of the set of possible disruptions 1 14 while meeting any constraints 220 of the base plan 1 16. if it is determined that the online time period 128 has not yet started at block 420, the incremental improvement of the set 212 at block 410 may be carried out continuously during an offline time period 1 18.

[0046] At block 420, if it is determined that the online time period 128 has started, the device 200, at block 430, selects a subset 122 of the set of recovery plans 212 that overcome the combination 214 of the possible disruptions closest to a set of actual disruptions 124. Then, at block 440, the device 200 further incrementally improves the solution for the subset 122 to reduce a cost 126 for at least one of the subset of recovery plans 122. At block 450,

[0047] if it is determined that the online time period 128 has not yet ended at block 450, the incremental improvement of the subset 122 at block 440 may be carried out continuously. However, if at block 450, if it is determined that the online time period 128 has ended, the device 200, at block 460, selects the recovery plan of the subset 122 having the lowest cost 126.

[0048] The cost 126 of each of the recovery plans 122 may be based on consequences of a path of the recovery plan 122 to overcome the corresponding combination 214 of the set of possible disruptions 1 14. An order in which the recovery plans of the subset 122 are chosen to be incrementally improved, may be based on the costs 126 of the recovery plans of the subset 122.

Claims

CLASMS We claim:

1 . A device, comprising:

an offline unit to calculate and store a set of recovery plans based on a set of possible disruptions to a base plan and to incrementally improve solutions to the set of recovery plans during an offline time period, each of the set of recovery plans to be calculated for a combination of the set of possible disruptions; and

an online unit to select a subset of the stored set of recovery plans during an online time period after the offline time period, based on a match between a set of actual disruptions and the set of possible disruptions, wherein

the online unit is to continue to incrementally improve at least one of the subset during the online time period, if at least one of the subset has a cost that is reducible, and

the online unit is to select one of the recovery plans of the subset based on a cost calculated for the subset of recovery plans.

2. The device of claim 1 , wherein,

the online unit is to select the recovery plan having a lowest of the costs of the subset,

the online unit is to select the subset when at least one disruption is added to the set of actual disruptions,

at least one disruption is added to the set of actual disruptions when at least one disruption at least one of has occurred and will occur, and the online time period is significantly shorter than the offline time period.

3. The device of claim 1 , wherein,

the offline unit is to select an order to calculate the set of recovery plans based on which of the set of possible disruptions at least one of has a greater likelihood to occur and takes a greater time to re-plan around, and

the offline unit is to incrementally improve the set of recovery plans by iteratively re-planning the base plan such that the cost of the corresponding recovery plan is non-increasing as computation time increases.

4. The device of claim 1 , wherein,

the online unit is to select the subset of the set of recovery plans based a greatest number of the disruptions that match between the sets of possible and actual disruptions, and

the online unit is to stop incrementally re-planning the cost of at least one of the subset if the online time period expires.

5. The device of claim 1 , wherein,

the set of recovery plans includes alternate plans to recover from a breakdown of the base plan caused by the set of possible disruptions, and the set of recovery plans are to meet constraints of the base plan, and the offline unit is to incrementally calculate the set of recovery plans to have a lower cost.

6. The device of claim 1 , wherein,

the cost of each of the set of recovery plans is based on a cost function of the corresponding recovery plan, and

the cost function is to relate to a weighted sum of consequences caused by the recovery plan in response to the corresponding combination of the set of possible disruptions,

7. The device of claim 6, wherein,

the consequences having at least one of a greater likelihood to occur and greater penalty for re-planning are associated with a greater weight,

the offline unit is to incrementally reduce the cost functions of the set of possible recovery plans, and

the online unit is to incrementally further reduce the cost functions of the subset of recovery plans, if the cost functions of the subset of recovery plans are further reducible.

8. The device of claim 8, wherein,

the base plan relates to a set of flights to be operated by an airline over a given time period, and

at least one of the offline and online units incrementally reduce the cost functions by incrementally revising the set of flights while meeting constraints of the base plan and overcoming any of the disruptions of at least one of the set of possible and actual disruptions.

9. The device of claim 8, wherein,

the constraints are associated with at least of an aircraft, an airport, itineraries, turn-around times, and a crew

the set of possible disruptions include at least one of a flight, aircraft, airport, and crew disruption, and

the cost function relates to at least one of operating costs, passenger inconvenience costs, and inconsistency costs that are incurred if the positions of the aircraft at the end of a recovery period do not match a planned position.

10. The device of claim 9, wherein,

the constraints further relate to at least one of seating capacities, maintenance, airport capacities, minimum connection time, minimum turn-round time, and transit time,

the flight disruption relates to a flight being delayed or cancelled, the aircraft disruption relates to a time the aircraft is unavailable, and the airport disruption relates to temporarily reduced departure and arrival capacities of the airport.

1 1 . The device of claim 1 , wherein, the offline unit is to store the set of recovery plans at a repository including at least one of random-access memory device (RAM), non-volatile random-access memory (NVRAM), a flash drive, a disk drive and a network-connected storage element.

12. A computer-implemented method, comprising: improving incrementally solutions for a set of recovery plans based on a base plan and a set of possible disruptions, each of the set of recovery plans to provide an alternate solution to overcoming a combination of the set of possible disruptions while meeting any constraints of the base plan;

selecting a subset of the set of recovery plans that overcome the combination of the possible disruptions closest to a set of actual disruptions, if an online time period starts;

improving incrementally the solution for the subset, during the online time period, to reduce a cost for at least one of the subset of recovery plans; and selecting, at the end of the online time, the recovery plan of the subset having the lowest cost.

13. The computer-implemented method of claim 12, wherein, the cost of each of the recovery plans is based on consequences of a path of the recovery plan to overcome the corresponding combination of set of possible disruptions, and

an order that the recovery plans of the subset are incrementally improved is based on the costs of the recovery plans of the subset.

14. A non-transitory computer-readable storage medium storing instructions that, if executed by a processor of a device, cause the processor to: prepare a set of recovery plans based on a base plan and a set of possible disruptions, the set of recovery plans to re-plan the base plan to overcome combinations of the set of possible disruptions; incrementally reduce a cost of the set of recovery plans during an offline time period;

select a subset of the set of recovery plans having a closest match to a set of actual disruptions, the closet match to be based on at least one of a number and importance of any of the disruptions matching between the set of possible and actual disruptions;

incrementally further reduce the cost of at least one of the subset of recovery plans during an online time period after the offline time period, the cost to be based on consequences of the recovery plan to overcome the corresponding combination of possible disruptions; and

select the recovery plan of the subset of having a lowest of the costs.

15. The non-transitory computer-readable storage medium of claim 14, wherein the cost is incrementally calculated by iteratively swapping pairs of elements of the recovery plan while meeting a constraint of the base plan and continuing to overcome the corresponding combination of possible disruptions.