1 / 23

Innovations in Automated Planning and Scheduling

Innovations in Automated Planning and Scheduling. 1st workshop of the EC AUTOMAIN Project Francis SOURD – SNCF – WP5 leader Paris, October 4th 2012. WP5 team. Objective of the work - Definitions.

freya
Download Presentation

Innovations in Automated Planning and Scheduling

An Image/Link below is provided (as is) to download presentation Download Policy: Content on the Website is provided to you AS IS for your information and personal use and may not be sold / licensed / shared on other websites without getting consent from its author. Content is provided to you AS IS for your information and personal use only. Download presentation by click this link. While downloading, if for some reason you are not able to download a presentation, the publisher may have deleted the file from their server. During download, if you can't get a presentation, the file might be deleted by the publisher.

E N D

Presentation Transcript


  1. Innovations in Automated Planning and Scheduling 1st workshop of the EC AUTOMAIN Project Francis SOURD – SNCF – WP5 leader Paris, October 4th 2012 www.automain.eu A Joint Research Project funded under the Seventh Framework Programme (FP7) of the European Commission

  2. WP5 team

  3. Objective of the work - Definitions • Develop operations research methods and tools for autonomous maintenance planning and scheduling. • Planning identifies the time periods when a track segment should be closed for maintenance. • Scheduling computes the start and end times of the operations and adapt the timetables of the commercial trains.

  4. Operations Research • Operations Research: application of advanced analytical methods to help make better decisions • Here (as often) the advanced analytical methods are mathematical optimization methods in order to automatically compute • the best feasible solution • or at least some optimized good schedules.

  5. Operations Research approach • What is a solution? • Fixed parameters.  KNOWN • length of track segment, maintenance operations… • Decision variables.  UNKNOWN • A solution is defined (non ambiguously) when the decision variables are instantiated (values are assigned all the variables). • Start dates and times

  6. Operations Research approach • What is a feasible solution? • List the constraints that a planning (or a schedule) must satisfy • Express these constraints as a mathematical (in)equality in function of the decision variables. • If all the (in)equality are satisfied when the decision variables are instantiated, the solution is feasible.

  7. Operations Research approach • What is a good solution? • Introduce a mathematical function depending of the decision variables  the objective function • For each solution, that is for each instantiation of the decision variables, the objective function can be evaluated • The higher the evaluation is, the better the solution. •  Maximize the objective function with respect to the constraints.

  8. Operations Research approachSome practical considerations • We must be able to feed the model with good numerical values for the parameters. • Some constraints may be violated. • Some constraints are missing. • There is no unique objective function. • Optimisation is complex and takes CPU time.

  9. Collaborative planning system Data Interface language CollaborativeplanningsystemConflict detection & GUI Data Interface language OR module OR module Maintenanceneeds OR module OR module

  10. Definition of the problems • Work in relation with WP1 • Two sources • Analysis of the answers to the questionnaire • Analysis of the state-of-the-art • Four new models introduced • Long-term planning problem (LTPP) • Dynamic planning problem (DynPP) • Time-window insertion problem (TWIP) • Work Site Scheduling problem (WSSP)

  11. Long-term planning (LTP) • Finds the best days to execute the maintenance operations (daily planning) • Planning over several years (typically 3 years)

  12. LTP constraints • Operation combination constraints • defined for maintenance types • Routing constraints for maintenance/inspection machines • Algorithmic collaboration with TWIP (via TWG) • Track availability constraints (not yet implemented) • Macroscopic description • Maximum possession time for a segment • Maximum possession time for a sub-network (set of segments) • Maximum number of possessions • Incompatibility constraints between track possession • The simultaneous possession of two track segments can be forbidden in order to continue the service between two points of the network.

  13. LTP Objectives • Minimization of track possession for inspection, maintenance and moves of machines • Minimize the total cost • Maximization of the use of maintenance machines • Number of required maintenance machine • Work load balancing between pre-determined sub-networks (not yet implemented). • work to improve the model is necessary

  14. Dynamic planningNot yet implemented • Variant of LTP  re-planning • The long-term planning is given in input • Some additional maintenance operations become necessary after inspection • They must be inserted in the planning/schedule • Minimize the insertion cost • Minimize the impact of these new tasks on the initial planning (update cost)

  15. Time window insertion problem(TWIP) • Input • A railway network • A fixed schedule for commercial freight and passenger trains • Over about 24 hours • Time must be limited due to computational complexity. • A short list of time windows and logistics or inspection train paths to be inserted in the commercial schedule

  16. Simple time window insertion(Example) km E’ E=S’ S t

  17. TWIP constraints • No conflict between paths is allowed, may they be technical or commercial. • An input path or time window can be “deformable”: • Maintenance train can be parked for some time in some predefined points • Speed of the maintenance train is subject to a minimal and maximal speed • Some time windows could be defined with alternative modes • for instance, 1 single window of 2 hours or 2 windows of 1.5 hours • Generalized temporal constraints • arrival of the technical train at the latest 30 minutes after the beginning of the works • All the paths and windows must be inserted • Indeed, the paths and windows given in input are related to each other. We assume that they are all required to perform the maintenance task. • Their number is not too large.

  18. TWIP Objectives • Minimize the cost • a cost function must be given in order to assess the cost of a time window according to its start time and its duration • Minimize the duration • The duration is the time span between the start time of the earliest time window and the completion of the latest one. • For example, if one train path is to be inserted, this objective function will minimize the total stopping time of the train • Minimize the disturbances on the business service • If it is not possible to insert the operations without modifying the business service, a degraded mode can be considered, with the possibility to delay, advance or remove trains. • Penalties for early, late and cancelled trains must be given in input.

  19. Work site scheduling problem(Not implemented) • Variant of TWIP • Shorter time span and smaller sub-network • typically the time and space extent of a track possession • More objects to be inserted • Here a time window corresponds to a basic maintenance operation • Advanced compatibility constraints are required • Resource constraints • Track / security constraints

  20. Solution approachCollaborative optimization Reference Data XML-based file format defined - RailML import not supported in D5.1 LTP module Macroscopic long-term planning TWG module Time-window and train paths generation TWIP module Microscopic time-window scheduling

  21. The three problems in the tool • Long-term planning problem (LTP) • Large scale (whole country) over 1-3 years • Resource requirements and capacities • Time-window/track possession generation (TWG) • Cost-time trade-off for moving a maintenance machine • Cost-time trade-off for performing a maintenance operations • Time-window insertion problem (TWIP) • Given existing train paths, how to insert the track possessions in the timetable (local scale, over a few hours)

  22. Work flow – D5.1 PDD TWIP – WSPP LTP – DynP Informal description All Development LTP - SNCF D5.1 Prototype ED+SNCF+TUBS Test instances All PSD LTP TWG  TWIP Formal models Software architecture Algorithms ED + SNCF + TUBS D5.2 Demo All Development TWG - TUBS Development TWIP - ED Maintenance data MERMEC – WP3 ? Completed Network - Trains SNCF – NR/WP3? Running Not started GUI implementation ? GUI – MMI in WP3? DLR Problem

  23. Next steps • Module development phase is finishing. • Test case is about to be released. • Test and Integration phase in October – December. • Release of D5.1 (beta version) in January 2013. • Tool will then be finalized. • Experimental tests will compare different scenarios based on other WP results.

More Related