In this chapter, we carry out an overview and analysis of the usage of semantics to enhance environments in the domain of multiagent-based simulations. Firstly, we take a look at what a multiagent system (MAS) is, and after that we look at the environment for these systems, and why semantics are required in it. Various propositions to put semantics in the environment for MAS are then reviewed, as well as the strengths and weaknesses for these approaches. These propositions are grouped together under two categories, regarding whether the proposed approach is based on only the environment or on both the agents and the environment. The paper is then concluded with findings that have emerged by analyzing the various proposed approaches.
TopIntroduction
According to (Ferber, 1995), six elements constitute a multiagent system (MAS): an environment; active objects (agents), passive object, relationships between objects, operations modeling the behavior of the agents, and the behavior of the environment.
An agent is an autonomous entity which is capable to act on itself or on its environment. It communicates with other agents, and its behavior is the result of its views, its knowledge and interactions with other agents. Multiagent-based simulations (MABS) solve problems that cannot be solved by an individual agent or a monolithic system. In MAS, situated agents are mentioned when they are located in an environment. The environment is a key point in MABS. (Weyns, Ominici, & Odell, 2007) propose three points of view on the environment:
- 1.
The part of the system which is outside the agents’ community,
- 2.
The coordination medium among the agents,
- 3.
The running platform.
The physical environment provides and manages the laws, rules, constraints and other policies that govern and support the physical “existence” of agents and other entities. To carry out properly an environment for the MABS, three main points have to be considered: the topological and geometrical description, the dynamics of the environment, and its semantics. The first two points are usually efficiently addressed by the various simulation systems that are available. However, it is not the case for the semantic that form a problem to solve to obtain realistic simulations. This chapter is a survey on the use of semantic to define the “meaning” of an environment.
TopBackground On Semantic In Environments
Semantic Virtual Environment (SVE) is used to enrich the information by semantic data, which cannot be deduced from the geometry (walkway, roadway, etc.) It enhances the interactions between agents and objects in the environment (Otto & Berlin, 2005). Indeed, it provides information on the environment (color of traffic light), including the possible actions that may be carried out on the environment.
The semantics in virtual environments can occur at three levels (Tutenel, Bidarra, Smelik, & Kraker, 2008). The lower level is dedicated to the objects. An object contains a number of physical and functional properties. For example, the opening state of a door is provided to an agent to decide to do. At a higher level, the relationships between different objects are specified: inclusion, proximity, coordination...
The formalism that has emerged for modeling the semantic information is the ontology (Troyer, Bille, Romero, & Stuer, 2003) (Otto & Berlin, 2005) (Pellens & De Troyer, 2005). They represent the knowledge embedded in the environment. For example, (Chu & Li, 2008) describe the semantics of the objects in a virtual environment with ontology, so that users can design their own animation procedures, and facilitating the agents’ path planning.
Several approaches propose to separate the representation of the environment, and the action-selection modules of the agents (Farenc, Boulic, & Thalmann, 1999) (Marwan Badawi, 2004) (Abaci, Ciger, & Thalmann, 2005) (Grimaldo, Lozano, Barber, & Vigueras, 2008). The two following sections are dedicated to these two parts.