| Item | Explanation |
| Trans. s. |
This determines the translation noise in the
transition model. The value is the standard deviation of the
Gaussian noise applied to the translation. The translation noise of
the robot's odometry is set to 1.0. For an efficient search for the
optima, the noise in the transition model should be higher than the
noise of the robot's odometry.
|
| Rot. s. |
This determines the rotational noise in the transition
model. The rotation noise of the robot's odometry is set to 0.04. |
| Sensor s. |
The uncertainty of the sensors used in the sensor model. The
'true' noise of the sensors is set to 1.0. |
| P(rand dist) |
The proportion of particles that are placed on a random position
in the map. This is applied every time step and can be used to
recover from a 'kidnap'. |
| Niching. |
The niching method used to preserve the diversity in the
particle population. If set to 'none', the particle population will
converge to one solution if there are ambiguous situations. Possible
niching methods are 'crowding', 'closest-of-the-words', 'sharing',
'frequency dependent selection using 20% of the particles',
'frequency dependent selection using 1 particle', and 'local
selection'. For the later, a threshold need to be set to determine
the carrying capacity (Loc.sel. thr). |
| # particles |
The number of particles in the population. This will not work
for the local selection niching method, since it has an adaptive
population size. In that case 'Loc.sel thr' needs to be set to
control the carrying capacity. |
| Kidnap |
The robot will be placed to a random location with a random
orientation without updating the odometry. The particle filter will
therefore not be aware of the kidnap. |
| World |
Three different environments are available: ambiguous,
non-ambiguous (similar, but with a symmetry braking element), and complex. |
| Start/Pause |
To start and pause the simulation |
| Restart |
To restart the simulation |