Organising the Complexity of Behaviour

In [1] we have developed a computational model for multiple goals. The agent behaviour is organised in families of competing goals (plans) for determining priorities of execution. This paper extends the previous model by providing a method for executing complex plans consisting of a number of goals and/or other sub-plans. The various dependencies among goals/plans are analysed and a mechanism is suggested which provides additional voting to the priority of a component goal. For the class of independent goals it is possible to interleave sub-goals for achieving better performance. A number of simulation experiments are described and results indicate a more flexible behaviour. We conclude with a summary of findings and a discussion for future work.