Prediction Time in Anticipatory Systems

How important is it to anticipate what will happen in the future? Is it better to anticipate far into the future or to focus on the next few seconds? We have investigated this question within a multi-agent framework where four simulated robots try to collect gold from two buildings without being seen by two patrolling guard agents. The success of the task depends fundamentally on an ability to anticipate. Rosen [3] proposed that “an anticipatory system is: [...] a system containing a predictive model of itself and/or its environment, which allows it to change state at an instant in accord with the model’s predictions pertaining to a latter instant”. Davidsson [1] used simulations to investigate the benefits of anticipation. In the experiments, the task of the agents were to pick up targets in a two dimension grid world in a particular order and by using a linearly quasi-anticipatory agent architecture, cooperation between the agent lead to decreased total time for fetching all target objects. In previous work we investigated the importance of anticipation in navigation task [2]. The results of this work show that a multi-robot system will benefit from anticipation compared to a system without anticipation. However, the models used to predict must have high precision, otherwise a reactive or a pure planning strategy will perform equally well as anticipation. The benefits of anticipation also depends on the task. A complex task will increase the usefulness of anticipation and in a simple task, the reactive or planning strategies could perform better then the anticipation strategies. In this paper we will study one additional variable that affects anticipation. By varying the time for how long the agents will predict, we will try to answer the question of how the prediction time influences the success of a task.