Distributed Simulation Modeling of Warehousing Operations M. Zhou, Y. J. Son, C. Chen, and Q. Zhang Center for System Modeling and Simulation Indiana State University Terre Haute, Indiana 47809

Intro to distributed simulation What is a “distributed simulation”? A super-simulation process that contains multiple sub-simulation programs that are independently executing and interacting with each other, and coordinated through the super-simulation process Why distributed simulation? Composite and complex nature of real world systems Needs for synchronizing multiple asynchronized software processes/systems Needs for simulating geographically distanced systems (e.g. military applications) Applications

Intro to Warehousing/DC operations Key operational processes in warehouse-based Distribution Centers (DCs): Truck-docking process: activities to prepare trucks for loading or offloading at a DC In-bound process: receiving operations, e.g. offloading, inspecting, sorting and storing. Out-bound process: shipping operations, e.g. picking, staging, sorting, packing, and loading Cross-docking: breaking bulk, mixing and directly shipping without storage

Intro to warehousing processes In real world these processes are considered running separately or “independently” The activities/operations in the processes are inter-related in terms of resource (e.g. sharing the same crew of labor and equipment) and flow logic (e.g. truck docking affects both inbound and outbound operation, inventory management also affects inbound (replenishment) and outbound flows)

Exemplary interactions Inbound process assigns resource (e.g. an offloading crew that involves labor and equipment) and starts sorting and put-away storage operation when truck-docking process seizes and finishes connecting to a dock-door When Inbound process finishes off-loading activity, the truck will release the door and leave the yard (a waiting or staging space) Outbound process initializes picking-sorting-packing jobs as a result of order-processing (an activity of inventory management) Interactions are triggered by the events that are the consequences of the activities belonging to different processes

Simulation of subsystems and overall system To reduce modeling complexity, we usually build simulation models separately for each subsystem to evaluate its performance with assumptions (and constraints) about the interfaces between the subsystems Subsystems usually assume different “forms” of entities defined to satisfy the different objectives of analysis__ the objectives for analyzing that subsystem

An example … In a CD/DVD distribution process, three subsystems, a location-based picking process, an automatic sorting process, and a wrapping/packing/shipping process are running separately but coordinated through the control of a sequence of events Location-based Picking process Automatic sorting process Wrapping and shipping process Solid arrows: material flow; dashed arrows: information flow

Distributed simulation In order to effectively and efficiently simulate the overall performance of a DC, we need to integrate the individual simulation models developed for the subsystems (legacy systems), i.e. coordinate their executions in a distributed environment to accomplish the analysis Distributed simulation provides a technology that enables us to integrate and coordinate the simulation models developed for each subsystems, namely, truck-docking, inbound, and outbound process so that the performance of the overall system (a distribution center) can be more realistically and accurately described and evaluated through simulation

A distributed simulation environment Architectural components: Legacy models/systems (HLA based federates) Coordination/manager model (HLA based Federation manager) Data repository Interfaces/adapters (transmission, receipt and internal updates to all FOM objects used by a federate) Tools (e.g. output analysis) Integration infrastructure/platform (e.g. RTI based on the HLA standard)

A distributed simulation environment Data integration - Communications between subsystems are realized via messages that trigger the events and transfer common data structures MessageName {Sender; Receiver; List_FOM Object Attributes}; For example: Request_IB {TD; IB; (T.LoadType, T.SkuList, T.SkuQty; T.RequestedTaskType, …)}

A distributed simulation environment Time coordination When Tsub(i) < Tsys and need to advance if (Tsys – Tsub(i) ) >  then Tsub(i) = Tsub(i) +  else Tsub(i) = Tsys When Tsub(i) = Tsys and system advance not completed Do nothing (i.e. halt the local simulation until the system finishes all the advancement requests of subsystems)

A distributed simulation environment Legacy Model1 Adapter Coordination model (Federation Manager) Data Repository Run Time Infrastructure and Communication network

A distributed simulation environment High Level Architecture (HLA): an architectural standard for integrating distributed simulation or multiple software processes, by the DoD of USA. Run Time Infrastructure (RTI): a software implementation of HLA specifications Federation Object Model (FOM): a common data definition created for the domain data shared across the federation; each federate has a FOM object that defines the elements of the FOM that it implements Developing implementation of HLA RTI and the interfaces between federates and federation manager is NOT a trivial task!

The focus of our research This research focuses on the modeling and evaluation of logistics distribution center (DC) operations via distributed simulation One objective is to build a set of conceptual models to describe the subsystems, their interactions, and the coordination required for the evaluation of the overall system Second objective is to construct a prototype to verify and validate the proposed distributed simulation for the target domain

Proposed work: conceptual modeling Building conceptual models for the subsystems: Truck-docking process Inbound process Developing legacy simulation models for the subsystems (e.g. using ARENA©) Modeling the interactions between the subsystems via Deterministic Finite State Automata (DFSA) DFSA graphical model (logic level) Procedural or algorithmic model (implementation level)

Conceptual models for subsystems Trucks arrive Register at gate Assign to doors Connect to the door Door available? Offload merchandise Wait in the yard Assign off-loaders Release truck No E Yes Truck-docking process

Conceptual models for subsystems Connect trucks Inspection at dock Reject and return load Off-load goods Accept loads? Count, stage merchandise Release truck 2 Sort/label products Store or C-D? Assign handler Cross-dock the goods No Update inventory E Yes Inbound Process

DFSA model for truck docking process 7 6 5 4 1 2 3 Door_available_IT Request_IB_TI Request_IB_ok_IT Connect_to_door Off-load_TI Offload_finished_IT Release_truck Free_door_TI T I O I = Incoming message O = Outgoing message T = Task to be performed

DFSA model for inbound process 7 6 5 4 1 2 3 Request_IB_TI Request_IB_ok_IT Off-load_TI Off-load Free_Door_TI Offload_finished_IT Release_Door Door_Available_IT T I O

Conceptual modeling Define the set of messages (data structure and objects involved …, these messages trigger the events) Define FOM objects FOM objects are defined in a public domain, i.e. these definitions are known by a number of software processes (simulation processes) Define these objects in a standard way, e.g. following the DMS, an implementation of HLA/RTI for manufacturing simulation

Conceptual modeling Implement DFSA graphical models (Coordination Model) Algorithm/procedures (pseudo coding) Object-oriented design and representation Design and development through a unified process with UML Implement the representations with Rhapsody C++/Java (Indiana State University)

Conceptual modeling Define interface between legacy models, federation manager and RTI platform: Selection of an existing HLA/RTI implementation Developing adaptation code between legacy systems and RTI interface Challenges: RTI interface is complex (about 120 methods  RTI Ambassador Interface and 40  Federate Ambassador Interface) Legacy systems provide very limited and inadequate mechanisms for integration Coordinated time management is required

Prototype: integration with DMS Distributed manufacturing simulation (DMS) is a software developed by NIST under MISSION Project (McLean 2000) DMS provides a distributed computing environment where non-simulation software applications are running and interacting with one or more simulation systems Using DMS can significantly reduce the effort required in developing distributed simulation applications (i.e. developing the interface)

Prototype: integration with DMS DMS Adapter Adapter Interface Federate Ambassador HLA Run Time Infrastructure RTI Object Model Legacy Simulation Adaptation Code

Architecture of a prototype Truck-dock model C-Adapter Inbound Model Outbound Coordination model (Federation Manager) DC Simulation Data Repository DMS Adapter/Interface and Communication network

This research is ongoing, and … If you are interested in collaborating with us, please contact Dr. Ming Zhou at: imming@isugw.indstate.edu OR call us at (812)-237-3983