Towards automated procurement via agentaware negotiation support

1

• Towards automated procurement via agent-aware
  negotiation support
• Andrea Giovannucci, Juan A. Rodríguez-Aguilar
• Antonio Reyes, Jesus Cerquides, Xavier Noria

Artificial Intelligence Research Institute
Ljubljana March 1st 2005
2
Agenda
Motivation Requirements Model
Implementation Demo
3
Motivation. Parts purchasing
FRONT SUSPENSION, FRONT WHEEL BEARING ACQUISITION
GOAL BUY PARTS TO PRODUCE 200 CARS
4
Motivation
Typical negotiation (sourcing) event in
industrial procurement
5
Motivation

• Multi-item, multi-unit, multi-attribute
  negotiations in industrial procurement pose
  serious challenges to buying agents when trying
  to determine the best set of providing agents
  offers.
• A buying agents decision involves a large
  variety of preferences expressing his business
  rules.
• Providers require to express their business rules
  over their offering.

6
Goal

• To provide a negotiation service for buying
  agents to help them determine the optimal bundle
  of offers based on a large variety of constraints
  and preferences.
• assistance to buyers in one-to-many negotiations
  and
• automated winner-determination in combinatorial
  auctions.
• To relieve buying agents with the burden of
  solving too hard a problem (NP problem) and
  concentrate on strategic issues.

7
Agenda
Motivation Requirements Model
Implementation Demo
8
Requirements Buyer side

• Negotiation over multiple items.
• Fuzzy expressiveness to compose demands(e.g.
  quantity requested per item lies within some
  range).
• Safety constraints. Establish minimum/maximum
  percentage of units per item that can be
  allocated to a single provider.
• Capacity constraints. Allocated units cannot
  excede providers capacities.
• Item constraints. Capability of imposing
  constraints on the values a given items
  attributes take on.
• Inter-item constraints. Capability of imposing
  relationship on different items attributes.

9
Requirements Provider side

• Multiple bids/offers per provider
• Offers expressed over quantity ranges in batch
  sizes (e.g. Provider P offers Buyer B from 100 to
  200 3-inches screws in 25-unit buckets)
• Offers over bundles of items
• Types of offers over bundles
• XOR. Exclusive offers that cannot be
  simultaneously accepted.
• AND. Useful for providers whose pricing expressed
  as a combination of basis price and volumen-based
  price (e.g. Provider Ps unit price is 2.5 and
  different discounts are applied depending on
  volume of required items 1-10 units (2), 10-99
  (3), 100-1000 (5)).
• Homogeneous offers that enforce buyers to select
  equal number of units per offer item.

10
Agenda
Motivation Goal Requirements Model Agent
Service Description Demo
11
Model

• Modelled as a combinatorial problem defined as
  the optimisation(maximisation or minimisation)
  of
• yj. (binary) decision variable on for the
  submitted bids
• 0wj1 degree of importance assigned by the buyer
  to item i-th
• V1, , ........ Vm bid valuation functions per
  item
• qij decision variable on the number of units
  selected from j-th offer for i-th item
• pij unitary prices per item
• ?ij ltdi1j,, d ikjgt bid values offered by j-th
  bid for i-th item
• Realised as a variation of MDKP
  (multi-dimensional knapsack problem).

12
Model
SIDE CONSTRAINTS
FORMALISATION

• Units allocated to each provider falls within his
  offer
• Allocated units per bid multiple of bids batch
• Aggregation of selected bids units lies within
  requested ranges of units
• Units allocated to a single provider do not
  exceed his capacity
• Percentage of units allocated to a single
  provider does not exceed safety constraints

13
Model
SIDE CONSTRAINTS
FORMALISATION

• Homogeneous combinatorial bids must be satisfied
• Providers per item must comply with saftey
  constraints
• AND bids must be satisfied
• XOR bids must be satisfied
• Intra-item constraints must be satisfied
• Inter-item constraints must be satisfied

14
Agenda
Motivation Requirements Model
Implementation Demo
15
Service Architecture
RFQ
RFQ
RFQ
RFQ
16
Service Architecture
SOLUTION
SOLUTION
PROBLEM
PROPOSE (BIDS)
PROPOSE (BIDS)
17
AUML Interaction protocol
IP-CFP
IP-RFQ
IP Request Solution
Protocols implemented as JADE behaviours
(extensions of the FSMBehaviour class)
IP-AWARD
18
Service Ontology (I)
RFQ
ProviderResponse
Buyers Constraints
Providers Constraints
19
Service Ontology (II)
Bid Solution
Problem
20
Implementation features

• All agents in the agency implemented in JADE
• FIPA as ACL (agent communication language)
• Two implementations of SOLVER
• ILOG CPLEX SOLVER
• MIP modeller based on GNU GLPK library
• Ontology editor Protegé2000
• Ontology generator The Beangenerator Protege2000
  plugin to generate ready-to-use Java classes

21
iBundler _at_ work
BUYER
TRANSLATOR
RFQ
ProviderResponse
22
iBundler _at_ work
TRANSLATOR
BUYER
Problem
Solution
23
Agenda
Motivation Goal Requirements Model Agent
Service Description Demo
24
Demo Parts acquisition
FRONT SUSPENSION, FRONT WHEEL BEARING
GOAL BUY PARTS TO PRODUCE 200 CARS
25
iBUNDLER DEMO
26
Demo Contract Allocation. Unconstrained RFQ
Ignoring business rules may lead to inefficient
allocations of products/services!!!
Unbalanced allocation
Unsafe allocation
Unsafe allocation
27
Demo Contract Allocation. Constrained RFQ
Balanced allocation
Safe allocation
Safe allocation
28
Demo Conclusion
iBundler helps buyers providers to reach better
agreeements
29
Summary and future works

• iBundler is an agent-aware negotiation service to
  help buying agents to determine the optimal
  bundle of offers based on a large variety of
  constraints and preferences. It provides
• assistance to buyers in one-to-many negotiations
  and
• automated winner-determination in combinatorial
  auctions.
• What happens if all constraints cannot be met?
• Empirical evaluation of the agentified service vs
  web service
• How to support bidders?

30
Thank you ... Any questions?