The 2D motion model outlined in class assumes the robot is a diff-drive robot, and handles all four possible errors in motion.
When new odometry information is available, we predict the a new distribution of particles by sampling from the motion model for each particle.
The change in pose for each particle is outlined here (links to a slide).
k_1, k_2, k_3, and k_4 are hyperparameters we need to tune. (Tracked in a separate issue: #28.) For this issue, just implement the math and make some initial hypothesis for the hyperparameters.
Essentially we're adding noise to the change in pose between the new odometry information and the previous odometry information and updating each particle with the noisy change in pose.
The 2D motion model outlined in class assumes the robot is a diff-drive robot, and handles all four possible errors in motion.
When new odometry information is available, we predict the a new distribution of particles by sampling from the motion model for each particle.
The change in pose for each particle is outlined here (links to a slide).
k_1, k_2, k_3, and k_4are hyperparameters we need to tune. (Tracked in a separate issue: #28.) For this issue, just implement the math and make some initial hypothesis for the hyperparameters.Essentially we're adding noise to the change in pose between the new odometry information and the previous odometry information and updating each particle with the noisy change in pose.