Things to try:
-
making the integrate odom multiply u by 1.1 or so to take into account communication delay
-
if there is a wall between the particle and the last good position, weight it lower
Things Tried
- averaging the particle the correct way [CHECK This Makes things work better]
- changing estimate method based on localization [Check, this localizes quickly]
- adjust particle weight so that they can't jump [CHECK]
Predictive position estimates -
t_tot = t_odom_2 - t_odom_1 - time between the two odometry measurements t_PF_start - time at beginning of particle filter t_PF_end - time at the end of the particle filter
Things that need to be fixed - Response to hitting a wall How close to a roadmap point do you have to be to consider it visited?