CORMAS
COMMON-POOL RESOURCES MANAGMENT
AND MULTI-AGENTS SYSTEMS
CORMAS - P. Bommel
What are we doing ?
Development of an agent-based simulation
platform (CORMAS) dedicated to the field of
natural resources management
Test of a companion modelling approach
about how to use these types of models
Training courses, knowledge transfers
CORMAS - P. Bommel
Resource management and environment
Social
Dynamics
Biological
Dynamics
Interactions
CORMAS - P. Bommel
A definition of complex systems
Definition:
 Complex system : a set of elements interacting among them and
with the outside :
• Distinction between the whole (the system) and the remainder
(environment/outer): reliance to the question
• Predominance of interactions : more than the sum of the parts.
Other characteristics:
 Descriptions at multiple levels (a minima: elements / system)
 Emergence
 Structures making
CORMAS - P. Bommel
Approaches of complex systems
Analytical : element by element (neo-classical economy,
plot, individual, etc.)
Holistic or systemic : global behaviour of the system
(macro-economy, statistics)
Constructivist : articulation between individual
behaviours of the elements (local) and the global
behaviour of the system (global).
Suitable to study ecosystems and sociosystems
 Challenge of complexity : situations far from equilibrium
 Intelligibility : intuitive description in terms of objects and agents
rather than variables and equations
CORMAS - P. Bommel
Approaches of complex systems (2)
Example: fish reproduction
 Analytical :
• Behaviour of one fish
 Holistic :
• Schaefer model
 Constructivist :
• Each entity of the system is represented (Molecule, cell, plant,
animal, group, etc.).
• Reproductive behaviour => evolution of the population
• Interactions between entities (Modification, creation,
destruction)
• Local interactions (reproduction with neighbours) => spatially-explicit
model
CORMAS - P. Bommel
Why consider the individual?
Principle of individual uniqueness
 Mathematical models assume that individuals are
indifferent and interchangeable (exchange of two
individuals randomly chosen in the population)
 Weakness of the mixing hypothesis: the differences
between individuals drive evolution of systems
Principle of located interactions
 An organism is mainly affected by the other organisms
and by the environmental conditions in its spatiotemporal neighbourhood
CORMAS - P. Bommel
Why consider the individual?
CORMAS - P. Bommel
The essence of complexity
Interactions and viewpoints
CORMAS - P. Bommel
Complexity
and resources management
CORMAS - P. Bommel
Postulates
Individual and collective processes
dealing with natural resources
management are based on:
 Shared representations of interactions between
stakeholders and the ecosystem.
Towards modelling to :
 Articulate multiple viewpoints ;
 Articulate multiple levels (from local to global) ;
 Allow retrospective and prospective analyses ;
 Be understandable by the stakeholders.
CORMAS - P. Bommel
Modelling
and collective dynamics
?
?
?
Environment
(common-pool
resources)
?
CORMAS - P. Bommel
Shared representations of interactions between
stakeholders and the ecosystem
Environment
(common-pool
resources)
?
CORMAS - P. Bommel
An explicit representation of space
Some elements from CORMAS facilities
carryingCapacity=2 grid Size: 50x50
Initial nbOfPreys=1000 Initial nbOfPredators=70
3500
500
450
3000
400
2500
350
300
2000
250
nbPreys
nbPredators
1500
200
150
1000
100
500
50
0
0
200
400
600
800
1000
1200
1400
1600
1800
0
2000
CORMAS - P. Bommel
CORMAS: an agent-based simulation framework
Origin: the aggregation of MA models
experiments dedicated to the domain of
natural-resources management
Cormas is an agent-based simulation
framework
It is oriented towards the building of
simulation models
 a programming environment.
 It provides a framework for developing simulation models of
coordination modes between individuals and groups that
jointly exploit common resources.
CORMAS - P. Bommel
CORMAS web site
Common-pool Resources and Multi-Agent Systems
http://cormas.cirad.fr
CORMAS - P. Bommel
Spatialised multi-agent system
Agents using or
managing
resources
Spatial objects:
points of view
Environment
containing
resources, a
topographic
support
DB, GIS
Cellular automata
CORMAS
- P. Bommel
Each agent builds its own representation of the
environment
Spatial grid
DB, GIS
Cellular automata
CORMAS
- P. Bommel
Spatial hierarchy levels
Various ways of segmenting the space
Relations of composition between spatial
entities can define several hierarchical
levels
On which entity to associate the
processes of the dynamics of the
landscape ?
CORMAS - P. Bommel
Spatial
hierarchy
levels
CORMAS - P. Bommel
The basic level
The Cell = the spatial entity element
The grid = a network of automatons
CORMAS - P. Bommel
The basic level: Regular tessellation
Van
Neumann
Moore
CORMAS - P. Bommel
The basic level: Regular or Irregular tessellation
 From GIS (Raster and Vectorial mode)
CORMAS - P. Bommel
The basic level: Regular tessellation
 From GIS data (Raster mode)
CORMAS - P. Bommel
The basic level: Vectorial mode
CORMAS - P. Bommel
The aggregation level
Spatial entities as agents’ viewpoints
Aggregation as reification of POV
Example of aggregation with minimum size
CORMAS - P. Bommel
Representation
of a
Mediterranean
forest
Forester’s
viewpoint
Herb
Shrub
Tree
Rock
Shepherd’s
viewpoint
CORMAS - P. Bommel
Same hierarchy structure for polygonal spatial
entities (vectorial mode)
Elementary entities
Cultivated plots
CORMAS - P. Bommel
Some dynamics are strongly related to a specific
hierarchical level
Agricultural dynamics
defined at the level of the plots
Ecological dynamics
defined at the basic level
CORMAS - P. Bommel
Spatial object dynamics
generate
produce
Natural processes :
growth, dissemination of
vegetation
Shapes
of
spatial
objects
influence
Human strategies:
pasture, grubbing, clearing
modify
spatial dynamics
 Spatial index calculation
CORMAS - P. Bommel
Spatial strategies
Level 0
strategies
Check
Clear max
Brushwood
Let Nature works
Fire-break
shepherd
Landscape
Level 3
strategies
Biodiversity
Level 1
strategies
tourism’s
professional
naturalist
forester
farmer
shepherd
forester
Level 2 strategies
compact
grassland
naturalist
Protect
grassland
Biggest forest
CORMAS - P. Bommel
Exemple: Pursuit
carryingCapacity=2 grid Size: 50x50
Initial nbOfPreys=1000 Initial nbOfPredators=70
3500
500
450
3000
400
2500
350
300
2000
250
nbPreys
nbPredators
1500
200
150
1000
100
500
50
0
0
200
400
600
800
1000
1200
1400
1600
1800
0
2000
CORMAS - P. Bommel
Applications
Standard models
 Game of life (Conway)
 ECEC: Evolution of Cooperation (Pepper and Smuts)
 SPD (Nowak and May)
 SugarScape (Axtell and Epstein)
CORMAS - P. Bommel
Applications
Applied models
 [AWARE] : Agent-based Watershed Analyses for Resource and Economic
Sustainability in South Africa (Farolfi).
 [AutomateVote] : electoral ballot
 [Bohol] : Natural Resource Management of the Municipality of Loon in Bohol,
Philippines (Campo).
 [BrouteLaForêt] : spatial representations and interactions between individuals,
space and society (Bonnefoy).
 [Burkina] : soil quality indicators in Burkina Faso (Guillobez).
 [CatchScape] : River bassins management in north Thaïland (Becu, Perez,
Walker)
 [Didy] : multiple uses of a forest ecosystem in Madagascar (Abrami).
 [Djemiong] : hunting of wild meat in Cameroun (Le Page, Bousquet and
Bakam).
 [Dricol] : emergence of resource-sharing conventions (Thébaud and Locatelli).
 [Echos] : Economic behaviour analysis of the "Stockbreeding wastewater
system" actors at the Reunion Island (Farolfi, Bommel).
CORMAS - P. Bommel
Applications
Applied models
 [FauconColombe] : game theory and prey-predator model (Valeix).
 [FiliereRaphia] : raphia marketing system in Madagascar (Herimandimby,
Randriarijaona, Bousquet and Antona).
 [ForPast] : spatial transformations dynamics of sylvopastoral systems (Lardon and
Bommel).
 [Gemace] : multiple uses of wetlands in Camargue, France (Mathevet).
 [JLB] : spatial transformations dynamics of forest systems (Bonnefoy).
 [JuMel] : economic exchanges and emerging organizations (Rouchier).
 [Kayanza] : firewood in Burundi (Guizol, Ndikumadengue, Bousquet and Antona).
 [MagmaS] : exchange of stock-farm effluents in Reunion island (Martin, Piquet, Le
Page and Guerrin).
 [Markets] : Assessing the performance of different market institutions in West
Africa according to communication systems (Galtier).
 [Mejan] : pine encroachment of natural ecosystems in Lozère, South of France
(Etienne and Le Page).
 [Mobe] : regulation of firewood marketing systems in Niger (Martine Antona).
CORMAS - P. Bommel
Applications
Applied models
 [Nong Chok] : Land use change in a peri-urban area, Bangkok Thailand (Anwar
and Borne).
 [Orizi] : Small irrigation systems under free management (Perez and Becu).
 [Pasteur] : sparse resource sharing by herds in sahelian area (Bah and d'Aquino).
 [PlotsRental] : plot renting by individual contracts or by centralized auction system
(Bousquet and Le Page).
 [Potlatch] : economic exchanges and emerging organizations (Rouchier).
 [Sabah] : plantation development among small farmers in Malaysia (Guizol).
 [SaintGeorges] : pasture and overgrowing brushwood in a village of Lozère,
France (Lieurain).
 [Samba] : land use in North VietNam (Boissau, Jean-Castella).
 [SavaneAgents] : landscape dynamics, agent-based version (Gautier and
Bousquet).
 [SeaLab] : homing-like reproductive strategies (Le Page).
 [Sinuse] : distributed interactions between an underwater table and its users
(Feuillette).
CORMAS - P. Bommel
Applications
Applied models
 [SpatioDyn] : spatial dynamics modelling with GIS and MAS (Bonin and Le Page).
 [Spiders] : net building by social spiders, a model from Bourjot and Chevrier.
 [Stratagènes] : negotiation for phytogenetic resource local management in
Madagascar (Aubert and Le Page).
 [SylvoPast] : Sylvopastoral management and wildfire prevention in Mediterranean
forests (Etienne and LePage).
 [WsErosion] : soil erosion risk and agricultural diversification in a NorthernThailand watershed
 [Zambeze] : land-use dynamics in the Zambeze valley
CORMAS - P. Bommel
Future …
A community of 200 users
Interactive simulation => RPG and Cormas
Distributed interactive simulation
Towards a “Companion” Modelling Approach
CORMAS - P. Bommel
Models and users
Passive way :
 Simulation models are frequently used in a passive way, presenting
only the results of experiments performed with the model.
Sensitivity analysis:
 People who experience the system dynamics will yield a better
understanding of the model.
Interactive simulation :
 In a simulation game like Fishbanks (Meadows, 1989) the players make
decisions about fishing strategies and the computer computes the fish
catches. The players have only limited control over the environment.
A “Companion” Modelling Approach :
 Combining ABM and RpG
CORMAS - P. Bommel