Programming Cognitive Agents in Goal Draft c ©

Bibliography 119 Index 123 6 CONTENTS Preface The Goal language is a high-level programming language for programming multi-agent systems. The language has been designed for the programming of cognitive and autonomous decision-making agents. Goal is a rule-based language that supports reasoning about knowledge and the goals an agent pursues. Cognitive agents maintain a mental state and derive their decisions on what to do next from their beliefs and goals. This programming guide advocates a view of agent-oriented programming as programming with mental states. The language offers a rich and powerful set of programming constructs and features for writing agent programs. It also provides support for connecting multi-agent systems to environments such as simulators, games, and robots (by means of the EIS interface standard) and for running multiple cognitive agents on a distributed computing environment. Agents are usually connected to some environment where they control entities such as grippers for moving blocks, cars in traffic simulators, or bots in real-time games. The possibilities are endless and the Goal platform is distributed with a diverse set of environments for which agents can be programmed. This Programming Guide has been significantly updated compared to the previous version from 2012. The language syntax has been updated throughout and is explained in detail. A " Quick Start " chapter has been added that introduces the language in the first chapter for those who want to immediately start programming in the language. Several features that have been added to the language have also been incorporated into this version. One of the more important conceptual changes in this version of the Goal Programming Guide has been to move away from the concept of a rational agent to that of a cognitive agent. A programmer may be interested in programming optimal or near-optimal agents and, of course, can do so. The programming language, however, does not enforce a programmer to write agent programs that are optimal in this sense. Rather, it provides a tool for programming cognitive agents in the sense that these are agents that maintain a mental state. Goal has been extensively used at the Delft University of Technology in education and, besides Delft, has been used in education in many other places at both the Bachelor and Master level. Educational materials are available and can be requested from the author. Several assignments have been developed over time that ask students to program agents for simple …