Home | Tasks | Help | Blog | Register | Login

Maze

Level: tricky

Navigate your agent through a random maze and touch the five goals as quickly as possible.

For general GoiD concepts visit the help page. For more information on this task, read on.

The agent in this task is a small "robot" that starts off somewhere near the middle of a maze, facing a random direction. The agent's primary senses are two sensors (blue circles) on its left and right called "eyes" which are triggered when they intersect the wall, and a "probe" (also blue) which senses the distance of the wall directly in front of the robot. Secondarily, the agent is given constant indicator of the distance to its current goal (the yellow patch of ground).

There are four goal areas located in the corners of the maze, but only one will be active to start. When that one is touched, one of the other three will activate. The order of the corner-goals is random, so you'll need to rely heavily on the goalDistance sense when navigating. Once all four corner goals have been reached, the final goal will activate at the agent's starting location.

Though the maze is psuedo-random, there are a few important constants. The maze is contiguous, which means there are no islands, and it is possible to reach any point in the maze by following one of the walls. The maze is square, and more specifically, it's 21 wall-widths wide. The paths are the same widths as the walls. All angles are right angles. The entire square is always used. Using these facts, it should be possible to build an acurate representation of the space in a short period of time.

Unlike in other tasks, the maze robot is unable to sense any absolute values for position or direction, so must build a picture of its surroundings as it wanders. The probe is probably the most useful sense for this, but don't rely on it too heavily for navigation. Because the agent is round and the probe acts in a line, it's easy to run into corners without the probe sensing them. Integrate the agent's senses and build an acurate model of the maze to get the best time!

senses

• blocked – (boolean) true if the last movement attempt was prevent by a wall
• leftEye – (boolean) true if there is an object within the left senor's field
• rightEye – (boolean) same as leftEye, but for the right sensor
• probe – (double) distance to the wall directly in front of the agent (maximum = 40)
• goalDistance – (double) distance from the agent to the goal that is currently active

actuators

• forwardMovement (0 to 1.0)
• backwardMovement (0 to 0.5)
• turn (-0.5 to 0.5)

Play now: (or use the applet)

All tasks can be played either as applets or with Java Web Start. Applets are convenient, but the rendering in Web Start is much better. Web start is recommended.