Professional Documents
Culture Documents
Particle Filtering For Tracking and Localization: Frank Dellaert, Fall 07
Particle Filtering For Tracking and Localization: Frank Dellaert, Fall 07
Frank Dellaert
October 07
Bayesian Inference
Data
Belief before
Belief after
Frank Dellaert
October 07
World Knowledge
z
d
x
Sensor Model
P(z|x)
d
MostMonte Carlo Methods Tutorial : often analytic expression, can be learned Frank Dellaert October 07
P(x|z) ~ L(x;z)P(x)
data = z
model L(x;z)
Belief before = P(x) Prior Distribution of x
Tutorial : Monte Carlo Methods
Prior P(x)
Walls and obstacles out of range Sonar and laser have problems Horizontally mounted sensors have problems
Frank Dellaert
October 07
Frank Dellaert
October 07
Solution: Ceiling Camera Upward looking camera Model of the world = Ceiling Mosaic
Frank Dellaert
October 07
Frank Dellaert
October 07
Ceiling Mosaic
Tutorial : Monte Carlo Methods Frank Dellaert October 07
P(z|x) z h(x)
Frank Dellaert
October 07
Under Light
Next to Light
Elsewhere
Gaussian centered around mean x,y Mixture of Gaussians Finite element i.e. histogram Does not scale to large state spaces encountered in computer vision & robotics
Frank Dellaert
October 07
ABCDE
Tutorial : Monte Carlo Methods Frank Dellaert October 07
Frank Dellaert
October 07
motion
motion
Frank Dellaert
October 07
x2
Markov Localization
Fine discretization over {x,y,theta} Very successful: Rhino, Minerva, Xavier
Frank Dellaert
October 07
Frank Dellaert
October 07
Sampling as Representation
p(x)
x2
x1
Tutorial : Monte Carlo Methods Frank Dellaert October 07
Sampling Advantages
Arbitrary densities Memory = O(#samples) Only in Typical Set Great visualization tool !
minus: Approximate
Frank Dellaert
October 07
Mean
Variance (1D)
Frank Dellaert
October 07
Frank Dellaert
October 07
How to Sample ?
Target Density (x) Assumption: we can evaluate (x) up to an arbitrary multiplicative constant
How to Sample ?
Numerical Recipes in C, Chapter 7 Transformation method: Gaussians etc Rejection sampling Importance sampling
Frank Dellaert
October 07
Rejection Sampling
Target Density (x) Proposal Density q(x) and q need only be known up to a factor
(x)
q(x)
Image by MacKay
Tutorial : Monte Carlo Methods Frank Dellaert October 07
The Good
9% Rejection Rate
Frank Dellaert
October 07
the Bad
October 07
October 07
Frank Dellaert
October 07
Importance Sampling
Idea:
sample from P(x) give each sample x(r) a importance weight equal to P(Z|x (r))
Tutorial : Monte Carlo Methods Frank Dellaert October 07
Frank Dellaert
October 07
Image by MacKay
Tutorial : Monte Carlo Methods Frank Dellaert October 07
1D Importance Sampling
Frank Dellaert
October 07
Example 1
Learn Color Model Implement Rejection Sampling Implement Importance Sampling Add a spatial prior
Frank Dellaert
October 07
Frank Dellaert
October 07
Likelihood
x
$ N( ,#
c c= r,g,b
^2)
October 07
References Isard & Blake 98, Condensation -- conditional density propagation for visual tracking Dellaert, Fox, Burgard & Thrun 99, Monte Carlo Methods Localization for Mobile Robots Khan, Balch & Dellaert 04 A RaoBlackwellized Particle Filter for EigenTracking
Frank Dellaert
October 07
Prior P(x)
1D Importance Sampling
Frank Dellaert
October 07
Measurement Zt-2
Zt-1
Zt
Frank Dellaert
October 07
Predictive Density
Frank Dellaert
October 07
Particle Filter
Empirical predictive density = Mixture Model
(1) (2)
(3)
Tutorial : Monte Carlo Methods Frank Dellaert October 07 First appeared in 70s, re-discovered by Kitagawa, Isard,
Sk-1
weighted
Sk
Sk
Predict
Tutorial : Monte Carlo Methods
Weight
Frank Dellaert
Resample
October 07
Frank Dellaert
October 07
Resampling
Frank Dellaert
October 07
Condensation Algorithm
Frank Dellaert
October 07
1. Prediction Phase
u P(xt| ,u)
Motion Model
Frank Dellaert
October 07
2. Measurement Phase
P(Z|xt)
Sensor Model
Tutorial : Monte Carlo Methods Frank Dellaert October 07
3. Resampling Step
O(N)
Frank Dellaert
October 07
Example 1
Start from Importance Sampling w Prior Implement Sample Mean Try increasing nrSamples Implement Resampling Step Implement Particle Motion Model
Frank Dellaert
October 07
Frank Dellaert
October 07
Likelihood
x
$ N( ,#
c c= r,g,b
^2)
October 07