In today's (2021-10-01) Lecture, Dr. Biswas outlined better ways to estimate the robot's location based on particles.
Weighted Mean
Clustering
Clustering can be more robust than finding the mean of all the particles. The example he gave in class was that the distribution of particles can be roughly bimodal; in this case the mean would find a value between the two modes with a potentially small weight.
Angle Mean
Finding the arithmetic mean of all the angles doesn't make sense since angles are not defined as reals. Instead, we can take the weighted average of the positions of the angles on the unit circle and then take atan2.
In today's (2021-10-01) Lecture, Dr. Biswas outlined better ways to estimate the robot's location based on particles.
Clustering can be more robust than finding the mean of all the particles. The example he gave in class was that the distribution of particles can be roughly bimodal; in this case the mean would find a value between the two modes with a potentially small weight.
Angle Mean
Finding the arithmetic mean of all the angles doesn't make sense since angles are not defined as reals. Instead, we can take the weighted average of the positions of the angles on the unit circle and then take
atan2.