Difference between revisions of "Trilateration in Robotic Sensing using Acoustic Conclusions"
| (6 intermediate revisions by the same user not shown) | |||
| Line 1: | Line 1: | ||
<sidebar>Trilateration_in_Robotic_Sensing_using_Acoustic_Sensors_Nav</sidebar> | <sidebar>Trilateration_in_Robotic_Sensing_using_Acoustic_Sensors_Nav</sidebar> | ||
===Conclusions and Future Work=== | ===Conclusions and Future Work=== | ||
| + | Our results this semester indicate that neither quantization nor noise should cause significant error with our experimental setup. | ||
| + | |||
| + | The performance analysis on the resolution map, or web, demonstrates that the resolution of points is extremely dense at a sampling frequency of 50kHz. This means quantization error will be minimal. Notably, the web includes all mathematically calculable points but many of those will not be used realistically. In the future we might attempt to create a web containing only those points which are physically viable. It would also be interesting to see how the arrangement of anchor nodes affects the shape of this web. | ||
| + | |||
| + | The simulations we ran on the effect of noise guarantee us that, with a high volume chirp relative to the surrounding noise, there will be little to no variance for meters away from a given speaker. The bias we came across, which is related to bandwidth, is something worthy of further investigation. Ideally it can be accounted for and subtracted to improve results. | ||
| + | |||
| + | Finally, all of these results have to do with locating a stationary target. Experiments run with a mobile robot demonstrated that our algorithm was not able to track a moving target, though we could locate the robot shortly after it came to rest. In the future we would like to enhance our algorithm to be able to locate an object continually while it moves. | ||
Latest revision as of 22:19, 21 December 2010
<sidebar>Trilateration_in_Robotic_Sensing_using_Acoustic_Sensors_Nav</sidebar>
Conclusions and Future Work
Our results this semester indicate that neither quantization nor noise should cause significant error with our experimental setup.
The performance analysis on the resolution map, or web, demonstrates that the resolution of points is extremely dense at a sampling frequency of 50kHz. This means quantization error will be minimal. Notably, the web includes all mathematically calculable points but many of those will not be used realistically. In the future we might attempt to create a web containing only those points which are physically viable. It would also be interesting to see how the arrangement of anchor nodes affects the shape of this web.
The simulations we ran on the effect of noise guarantee us that, with a high volume chirp relative to the surrounding noise, there will be little to no variance for meters away from a given speaker. The bias we came across, which is related to bandwidth, is something worthy of further investigation. Ideally it can be accounted for and subtracted to improve results.
Finally, all of these results have to do with locating a stationary target. Experiments run with a mobile robot demonstrated that our algorithm was not able to track a moving target, though we could locate the robot shortly after it came to rest. In the future we would like to enhance our algorithm to be able to locate an object continually while it moves.
