DOC PREVIEW
UT CS 395T - A General Equilibrium Based Price-Taking Trading Agent

This preview shows page 1-2-3-4 out of 11 pages.

Save
View full document
View full document
Premium Document
Do you want full access? Go Premium and unlock all 11 pages.
Access to all documents
Download any document
Ad free experience
View full document
Premium Document
Do you want full access? Go Premium and unlock all 11 pages.
Access to all documents
Download any document
Ad free experience
View full document
Premium Document
Do you want full access? Go Premium and unlock all 11 pages.
Access to all documents
Download any document
Ad free experience
View full document
Premium Document
Do you want full access? Go Premium and unlock all 11 pages.
Access to all documents
Download any document
Ad free experience
Premium Document
Do you want full access? Go Premium and unlock all 11 pages.
Access to all documents
Download any document
Ad free experience

Unformatted text preview:

Geep: A General Equilibrium BasedPrice-Taking Trading AgentAlan OurslandUniversity of Texas at AustinArtificial Intelligence LaboratoryAustin, TX 78712 USADecember 4, 2003AbstractGeep is an agent that competes in the Supply Chain Management series of theTrading Agent Competition. Geep attempts to achieve general equilibrium onthe customer side, and acts as a price taker on the supplier side. Geep uses ascheduling pipeline to manage its production capacity as effectively aspossible.1 IntroductionGeneral equilibrium is a stable market condition in which prices are set so that supplyequals demand. A price-taking buyer’s demand does not change based on the price. Geepis an agent that competes in the TAC Supply Chain Management competition. Geep’sstrategy is based on the idea of finding a general equilibrium solution on the customerside and acting as a price taker on the supplier side. The name Geep is derived from thisstrategy: GEneral Equilibrium Price-taker.2 General EquilibriumGeneral equilibrium consists of three conditions [1]:1. Consumption equals production (plus endowment).2. Each consumer maximizes its utility.3. Each producer maximizes its profits.In SCM this can be restated as: there is no surplus at the end of game; customers preferlower prices; and agents prefer to maximize profits.The different customers in SCM accept offers at different prices. This range in prices canbe described with in a demand curve. A specific amount of product will sell at each pricepoint. The agents need to find the price point that maximizes their profit. If the price isset beneath this point then there is a product shortage. If the price is set above this pointthen there is a product surplus.In most SCM games, demand exceeds the total agent production. Agents should try tofully utilize their production capacity and sell goods at as high a price as possible. Mostof the work in this version of Geep optimizes order processing and factory utilization.3 Geep ArchitectureGeep was designed around a scheduling pipeline. Integrated with the pipeline are acustomer request response mechanism, a component ordering mechanism, and a margincontroller mechanism.3.1 Pipeline OverviewThe pipeline consists of a set of plans across the days. Each day plan contains factoryusage; products in and out; components in and out; production schedules; and shippingschedules. The customer response mechanism can retrieve from the pipeline the numberof components and the factory time available on a given day. An offer is made on arequest only if it fits into the pipeline.The current implementation of the scheduling pipeline has one day of inefficiency.Orders sit in inventory for one day between production and shipping. This is an artifact ofwhere the in/out records for the products are kept. The “product in” records should reallybe on the previous day. However, this would lead to a disagreement with the inventoryrecords of the SCM library. There was not time to remove this inefficiency before thecontest, but it should be removed in a future version.Each day plan keeps track of:• Balance trackers for each of the components• Balance trackers for each of the products• The amount of unscheduled factory time for the day• The amount of factory time being reserved for the day• A list of orders to begin production on the day• A list of orders to ship on the dayThe balance trackers for each of the components and products keep track of:• Quantity coming in that day• Quantity going out that day• Quantity reserved for the day• Quantity that this day is borrowing from the previous day to make up for deficitsin this day• Quantity that is being borrowed from this day by the next day to make up fordeficits in the next dayReserved values are used during the RFQ response process. As offers are made tocustomers, the products, components, and factory time needed to fulfill the offer arereserved on the day. Offers are not extended if reserved resources exceed the availableresources. However, the resources are not actually marked as used until the offers areaccepted as orders. The reserved values are reset after all customer RFQ responses havebeen sent.On any given day n, one can determine the amount of some resource r that is availableusing the following formula:Availablen,r = Availablen-1,r + Inn,r – Outn,r – Reservedn,r + Borrowingn,r – Borrowedn,rSuppose an RFQ due on day ten is being considered. The system knows that this orderneeds to ship on day nine, and needs to start production on day six. It will first look at theamount of available factory time components on day six. If the resources required forproduction are available, the resources will be marked as reserved and offer will be madeon that RFQ.If a customer order is placed on that offer, then the order is added to the production liston day six and to the shipping list on day nine. The production time for the order will besubtracted from the day six available factory time. The out field for each of thecomponents used in manufacturing will be incremented by the quantity manufactured forday six. The product in for day nine will be incremented by the amount produced, and theproduct out on day nine will be incremented by the amount to ship. If the in field issmaller than the out plus the borrowed field, the day plan will borrow resources from theprevious day. The borrowing field in the current day and the borrowed field in theprevious day are both incremented as a result. If the previous day’s in is less than its newout + borrowed value, it will borrow resources from its previous day. This continuesuntil the first day is reached. When the in field of a resource in incremented, the samenumber is subtracted from the borrowing field which then percolates through theborrowed fields of the previous days. Each day has an expected value on each productand component. It knows how many it is producing, using, and how many are in use bylater days. Knowing this allows it to plan without exceeding the available resources.At the beginning of each day, all of the shipping and production commands stored in theprevious day plan are executed. Each of the executions is double-checked to make sure itis successful. Successful commands are deleted. Unsuccessful commands are moved tothe next day. The previous day plan is removed from the schedule pipeline. The inquantities on the earliest day reflect the current inventories. Available


View Full Document

UT CS 395T - A General Equilibrium Based Price-Taking Trading Agent

Documents in this Course
TERRA

TERRA

23 pages

OpenCL

OpenCL

15 pages

Byzantine

Byzantine

32 pages

Load more
Download A General Equilibrium Based Price-Taking Trading Agent
Our administrator received your request to download this document. We will send you the file to your email shortly.
Loading Unlocking...
Login

Join to view A General Equilibrium Based Price-Taking Trading Agent and access 3M+ class-specific study document.

or
We will never post anything without your permission.
Don't have an account?
Sign Up

Join to view A General Equilibrium Based Price-Taking Trading Agent 2 2 and access 3M+ class-specific study document.

or

By creating an account you agree to our Privacy Policy and Terms Of Use

Already a member?