2022 IB Diploma Extended Essays

Data from sensor-fusion Motion model

yaw i = yaw i-1 +w z i * dt i x i = x i-1 +v i * cos(yaw i ) * dt i y i = y i-1 + v i * sin(yaw i ) * dt i

Pose values for the Extended Kalman Filter Graph 1. Changes in x coordinates in relation to time

The first graph highlights the changes in x coordinates (m) in relation to time (seconds) for Hector, ORB and Zedfu Active SLAM algorithms and compares it to the Extended Kalman filter. The graph demonstrates that the LiDAR presents itself as the most robust passive slam algorithm. The maximum and minimum of each point highlights the robot changing it’s coordinates with the x plane in order to make a turn. While the smaller peaks represent the robot correcting it’s path through the change in the x position. The smoother the gradient of the curve’s the more accurate the robot is while doing autonomous navigation. In contrast to the LiDAR pose, ZEDfu presents more errors, while it generally follows the same shape as the LiDAR, there are some rough changes in gradient. Indicative of the robot needing to correct itself within the x coordinates. On the contrary ORB-SLAM presents quite a few errors within it’s path, while it takes less time to complete the maze, it needs to correct it’s position within the scene more frequently. But it still keeps the same