This extension aims to allow agent-based models to account for norms. During plan generation, agents must be able to represent and reason about norms. The end-goal is to be able to describe a planning problem with norms endowed by its organizations through the various roles that the agent must fulfill, and see how it affects its plans.
A few good lectures on how automated planning works and its core concepts :
Weld, D. S. (2011). AAAI-10 classic paper award: systematic nonlinear planning a commentary. AI Magazine, 32(1), 101-101.
Chapman, D. (1987). Planning for conjunctive goals. Artificial intelligence, 32(3), 333-377.
McAllester, D., & Rosenblatt, D. (1991). Systematic nonlinear planning.
This extension requires the following tools :
Java 11 or higher
Maven
Lombok
JUnit 5
The tool is based upon the Partial-Order Planner (POP), which refines a plan until it becomes executable and fulfill all the agent's goals. We start off by having a plan with just the initial and final situation, which will be improved by resolving its "flaws". In addition to the standard types of flaws, we propose three other types of flaws relative to regulativeNorms, and their resolvers.
Missing obligation
Missing prohibition
Missing permission
This will allow agents to have a clear mean to comply / circumvent / violate a norm.
To illustrate the type of problem this tool can solve, we propose a planning problelm for a single agent, named IRIELA, whose goal is to feed his family but is part of a lot of organizations : village.member, exploitation.farmer.
Its possible actions are defined by the organizations it is a part of :
- move
- cut
- fish
- getLicense
In the IRIELA planning problem, a norm MUST belong to some institution. It is then acquired by playing some role inside that institution by organizing which object plays which role and acquire which norms. This is called organization and is the core of our description about norms.
Using a simple, but verbose Kotlin-like syntax, all the institutions inside of the IRIELA planning problem.
To showcase a better grasp of how a planning problem might look like through our upcoming Domain-Specific Language (DSL) which allow to craft such planning problems, here is a snippet of the whole planning problem, including all those who acuquire roles and the norms from the aformentionned institutions.
Feel free to reach out to [email protected] or [email protected] if you have any feedback or comments on the current tool. Documentation is coming soon.