Using a Cognitive Architecture to Control the Behaviour of Virtual Robots.

摘要

In order to support the use of cognitive architectures in computer simulation studies involving virtual environments, an integration solution is proposed that enables ACT-R cognitive models to communicate with the Unity game engine. The integration solution is tested using virtual robots that are able to move around a 3D virtual environment and interact with various objects. The robots are equipped with visual, tactile and proprioceptive sensor systems, which yield information about the current sensory state of the robot. This information is communicated to an ACT-R model that subsequently controls the behaviour of the robot by making decisions about motor output. The performance of the model and the integrity of the integration solution are evaluated using a task that requires a virtual robot to retrieve target objects and deposit them in a specified goal location.