Action and Perception in Multi-Agent Programming Languages: From Exogenous to Endogenous Environments

The action and perception model adopted by current multi-agent programming languages has been conceived to work with exogenous environments, i.e. physical or even computational environments completely external to the multi-agent system (MAS) and then out of  MAS design and programming. In this paper we discuss the limits of adopting such models when endogenous environments are considered, i.e. fully computational (software) environments that are used by MAS developers as first-class abstraction in MAS engineering to be designed and programmed to encapsulate functionalities useful for, e.g., agent coordination, agent computational activities and agent access to the external environment. In the paper we describe an action and perception model for agent programming languages specifically conceived to be effective for  endogenous environments and we discuss its evaluation using {\cartago} environment technology. On the agent side,  we take Jason, 2APL and GOAL as reference agent programming languages.