next up previous
Next: Results Up: Testing and Results Previous: Free Ball

Accuracy

A second test was conducted to verify that the predictive memory was keeping an accurate model of the field. This was done by allowing the client to perform the same loose ball trial, but without the opposing player. The client would then dash for the ball unhindered, and kick it at the goal. The accuracy of the shot was then measured as the distance of the ball as it crossed the goal line from the center of the goal. For this experiment, the noise factor in the motion of the ball was removed, so that the raw accuracy of the memory model could be measured. It was expected that the simple memory would be more accurate since it is facing the goal when it shoots, while the predictive memory won't have seen the goal since it started dashing towards the ball. The results, though, were surprisingly close. They are shown in table 4.4. The results are given as a fraction of the size of the goal mouth.

 

  table68


: The tabulated results. Accuracy is given as the fraction of the size of the goal mouth.



Peter Stone
Wed Dec 17 13:44:25 EST 1997