2022 IB Diploma Extended Essays

lower minimums. Suggesting that it takes a wider angle of approach to make it’s turns. The Extended Kalman Filter makes no errors within the y plane, and presents a very smooth gradient. Therefore, showing that EKF is more robust at autonomous navigation. However, it can also be seen that Hector SLAM, shows the same results. This is because the robot manoeuvrability is constricted by Ackerman drive, this means that rotation of the robot occurs within the rear axle, as one wheel drives rotates slowly and the other rotates fast, changing direction. Therefore, it means that gradient in the y plane in relation to time will always be smooth. Showing little to no pose error, the smaller gradients would therefore show, the robot stopping and then moving. Graph 3. Changes in yaw in relation to time.

Graph 3. Highlights the changes in yaw rotational axis (radians) with relation to time (seconds). The incremental increase from 0 radians shows the robot front axle changing it’s angle in order to meet the path and therefore changing it’s orientation. This is expected as the robot needs to change it’s angle for the path. The flat sections represent the robot moving in a straight line, with no change in angle from the front axle. The smaller changes represent the robot correcting it’s angle (orientation) when it deviates from the path, due to pose estimation problems. The graph indicates that all SLAM algorithms have the same orientation. This would suggest that the path created from the teb_local_planner itself causes the robot too make quick changes in its orientation. This might be because the calculated path trajectory could have shown a tighter angle than the axle can change.