Projects

This research theme, within the framework of the COMECA project (CO‑creation of Innovative Forest Resources Management Combining Indigenous Knowledge and Ecological Methods), introduces a new approach to the design of cognitive agents that integrate spatial representation and reasoning into agent‑based modeling dedicated to land‑use simulations. The deep motivation behind our agent‑centered contribution lies in the growing development of spatially explicit agent‑based simulation platforms. We build upon this technological evolution and topology theory to endow agents with human‑like spatial representation and reasoning capacities, following a Belief–Desire–Intention (BDI) architecture.

A pilot implementation of the methodology, including simulation experiments on a hunting model, was carried out in the GAMA platform to assess agent performance. The simulations reveal a consistent trend in animal population dynamics and also confirm the model’s high sensitivity to the integration of spatial knowledge and reasoning in agent‑based models of human actors. These results demonstrate the successful implementation and highlight the importance of the spatial dimension in the expressive power and validity of agent‑based models.

Future research should therefore focus on designing representations of human knowledge and incorporating learning capabilities to improve model efficiency.

A Prototype Developed, Secured and Formally Validated for Institutional Adoption

TES currently serves four roles students, teachers, department deans, and platform administrators. each with a dedicated dashboard. Students submit anonymous course evaluations through a guided  form; teachers view aggregated results and comments for their own courses; deans and administrators access comparative statistics across departments and semesters, and export results as PDF reports, CSV data, or formatted Excel workbooks. Administrators additionally manage users, courses, evaluation criteria, semesters, and student ID cards, and can review a full audit trail of platform activity.

Three lines of evidence support the readiness claim developed through this report:
•  Functional  breadth:  the  full  workflow  (Chapter  3 of the main doc)  plus  18  additional  operational  scenarios (Chapters 4-6 of the main doc) were exercised live, role by role: creating a semester, adding and importing courses, students  and  teachers,  configuring  evaluation  questions,  enforcing  one-evaluation-per-student, isolating each teacher's data from every other, printing student ID cards singly and in batch, and generating  every  report  type  on  real  submitted  data.  Every  screenshot  in  this  document  was captured from those live runs.
•  Independent  quality  assurance :  three  audit  cycles  identified  and  resolved  concrete issues,  most  notably  strengthening  the  technical  guarantee  of  evaluation  anonymity  from  a reversible mechanism to an irreversible one, and catching a live regression in the login rate-limiter during the preparation of this very report.
•  Formal  verification  (Chapter  8 of the main doc):  a  peer-reviewable  mathematical  model  (Colored  Petri  Nets) proves  six  correctness  properties  of  the  platform's  core  workflows.

The growing complexity of socio‑environmental systems requires agent‑based models that capture human‑like cognition, particularly spatial reasoning. Existing GAMA architectures such as simple BDI and BEN represent deliberative agents with beliefs, desires, intentions, emotions, and norms, but they reason about space only implicitly through coordinates and distances. This limits realism in domains like land use, urban mobility, and crisis management.

We introduce SBDI (Spatial BDI), which embeds qualitative spatial reasoning based on the Region Connection Calculus (RCC8) into the belief base and deliberation cycle. RCC8 formalizes eight topological relations (DC, EC, PO, EQ, TPP, TPPI, NTPP, NTPPI), enabling agents to compute relations between their geometry and surrounding entities. Implemented as a lightweight GAML skill, SBDI extends simple BDI with three primitives: sbdi_perceive (detects RCC8 relations), sbdi_update (refreshes spatial knowledge), and sbdi_reason (triggers plans).

Validation through traffic simulations compared driver_basic, driver_bdi, and driver_sbdi populations. Results show SBDI drivers anticipate congestion via PO and EC relations, dynamically avoiding jams. Across 10 runs, congestion exposure dropped by ~60% (p < 1e‑14), with jam detections reduced from ~51,500 to ~19,800. Travel times remained statistically indistinguishable, confirming efficiency was preserved. Beyond traffic, SBDI offers reusable primitives for spatially explicit models in land use, logistics, and crisis response, with planned extensions for learning spatial predicates and validating temporal properties.

Research project

Irrespective of the field of study, a model is an abstract representation of the reality observed, scaled  down,  and  converted  to  a  form  that  is  easy  to  understand.  It  is  also  a  spatio-temporal reference for our understanding of a system [1]. Agent-based models (ABM) and spatial models in  Geographic  Information  System  (GIS)  have  proven  to  be  useful  tools  for  land  use  change processes analysis as they provide a framework to build models and conduct simulations.

On the one  hand,  GIS  is  a  set  of  computer-based  infrastructure,  human  resources,  procedures, and standards for the management of geographical information to address a specific issue [2]. GISbased models are composed of spatial entities including points, lines polygons,  or pixels. On the other hand, ABM allows to define the behavior of agents in a common environment for solving complex  problems.  It  is  also  used  to  model  a  distributed  computing  system  with  autonomous interacting agents that coordinate their actions to achieve their goal(s) jointly or competitively [3]. There are obvious similarities between agents and objects which usually prompt the designer of an ABM to rely on Unified Modeling Language formalism [4]. Both spatial and agent-based approaches  include  computer  simulation  platforms  often  used  in  several  research  fields (economics, ecology, sociology, geography, etc.) to develop models for explaining and analyzing complex phenomena [5], [6].

Several  authors  including  [7]–[10]  have  proposed  such  models  to  understand  the  drivers  of land  use  change  and  identify  appropriate  actions  to  maintain  the  target  system.  In interdisciplinary  contexts  and  participatory  modeling  situations,  there  is  an  increasing need to rapidly build more  realistic or effective models to solve environmental problems.

The key idea to achieve  this  is  to  rely  on  field  knowledge  to  mimic  the  attributes  and  behavior  of  the  entities involved. Different  research directions are investigated to increase the realism of the simulated environment. The first group of notable efforts oriented toward the integration of GIS and agentbased models can be  seen in [11],  [12]. Other efforts are oriented toward proposing appropriate agent architecture. In this  regard, exploring the concepts of  the  actor paradigm to build effective and  valid  agent  is  a  promising  research  pathway  [13],  [14].  In  addition  to  computer  and mathematical modeling (e.g. using ordinary differential equations [15]), agent design can rely on economic, social, and anthropological models [16].

The challenge of our research endeavor is to combine  actor  paradigm,  agent  architecture, and spatial concepts to develop effective models for addressing key environmental problems. The ad-hoc approach is  often  used  for  model  development,  which  relies  on  a  code-and-fix  life  cycle, going  from a prototype model through progressive refinement. However, this adaptive approach does not  capture systematically actors’  knowledge and their interactions with the environment.

Technically, the development  and  maintenance  of  resulting  models  become  cumbersome and time-consuming.  From these facts, the actor paradigm appears to  be a promising research focus in ABM as it links actor concepts to  those of an agent and therefore  improves the realism and validity  of  each  model  built.  Moreover,  to  cope  with  a  world  evolving  toward  more  applied artificial intelligence (AI), we need a more efficient and simple approach  as close as possible to fields actors  and appropriate for  non-computer scientists  that allows  to  capture  the reality, data, and  behavior  shared  within  an  interdisciplinary  community  and  then,  an  efficient  modeling technique to produce valid  ABM simulations. In this  article, we  propose an actor-oriented and architecture-driven  approach  called  Spatially  Explicit  Agent-based  Modeling  Approach (SEAMA) which relies on relevant existing methods and satisfies the need to design and build spatially explicit agent-based models that adequately mimic reality. We have designed an Agent Global Experiment (AGE) framework incorporating a meta-model built from actor, agent,  and spatial concepts to produce an initial model from specifications provided by domain experts  and system analysts.

Lien de l'article publié : click here

Image sur le terrain avec les habitants dans la régions de l'Est: