MASL: a Language for Multi-Agent System

Abstract

Our work takes place in the field of Multi-Agent Modular Robotic System. We propose to mix several paradigms of computation to offer a high-level point of view to the programmer into a new language, namely MASL for Multi-Agent System Language. Expressivity of MASL is illustrated on an example applied to a fleet of robots. A Multi-Robot System (MRS) can be characterized as a set of robots operating in the same environment that operate together to perform some global task. In this chapter, we regard MRS as a particular form of Multi-Agent System (MAS). The main differences between general MAS and MRS are: - The fact that in MRS direct communication is based on dedicated physical devices, resulting in a much more expensive and unreliable solution to attain coordination with respect to MAS. - The number of robots acting in the same environment is still quite limited with respect to the number of agents in a MAS. - Agents architectures in a MAS are usually deliberative (Nilsson, 1984) (Sense-Represent-Plan-Act), reactive (Brooks, 1989; Hudak et al., 2002) (subsumption architecture) or hybrid (Alur et al., 2000; Ingrand et al., 1996; Benjamin et al., 2004) as shown in figure 1 while in MRS we consider only the last two. The pure sense-represent-plan-act architecture, which is used to realize a high level deliberative behaviour, is not currently used in MRS because of its intrinsic limits, while the behaviour-based and the hybrid architectures are quite common, especially when the robot is situated in a highly dynamic environment, where a quick reaction to a new input is very important, being the environment itself uncertain and unpredictable. Robot programming is a difficult task that has been studied for many years (Lozano-Perez & Brooks, 1986). This particular field often covers some very different concepts such as methods or algorithms (planning, trajectory generation...). Therefore, languages are developed to implement these high-level considerations (Pembeci & Hager, 2001; Zielinski, 2000). Different approaches have appeared through functional (Armstrong, 1997; Atkin et al., 1999; King, 2002), deliberative or declarative (Dastani & van der Torre, 2003; Benjamin et al., 2004; Peterson et al., 1999), synchronous characteristics (Pembeci & Hager, 2001). Nonetheless, the difficulties of robot programming can by schematically summarized by two main characteristics: - One is that programming a set of elementary actions (primitives) on a robot often leads to (if not always) a program including many processes running in parallel with real-time constraints and local synchronizations.