Previous: The physical environment description Up: A simulation study: the Mobile Robot Lab Next: The graphical interface
The simulator interfaces with the agent and with the graphical interface. It takes care of any I/O with the agent that would otherwise come from the sensors and go to the actuators of a real mobile robot. It also takes care of any I/O with the graphical interface, needed to keep the graphical display of the robot and its physical environment updated.
The simulator incorporates a simplified model of the physics of motion and sensing for the mobile robot. It continually updates the position of the robot depending on the rotation speed and direction of its wheels, and provides the agent with appropriate sensory data about wheel rotation and contact with objects. It also prevents the robot from going ``through'' walls or objects. It provides simulated camera input to the agent. Camera input is simplified in that it consists of a 9x7 pixel array (square pixels), with each pixel represented as an RGB triplet. This simplified camera view is computed and passed to the simulator by the graphical interface, on the basis of the 3D perspective views (see below).
The simulator incorporates a simplified lighting model to determine the appearance (color) of objects in the room. Light sources can either be point sources or homogeneous diffuse sources. Each light source has its own SPD. Each object has its own spectral reflectance function. All objects are assumed to be Lambertian reflectors.
To enhance the realism of the simulation and to introduce some uncertainty into the environment, all robot controls and sensors work stochastically. Sensor readings vary stochastically over time around the ``true'' value, and the same is true for actuator output. We assume normal distributions for all such variations, with variable standard deviations.