Jethro's Braindump

Likelihood Field Model

tags: Map Matching

Key Idea

Project an individual sensor measurement $z_{t}^{k}$ into the global coordinate frame of map $m$ . Discards max-range readings.

Assumes three types of noise, similar to Range Finder Model:

Measurement noise: Gaussian
Failures: point-mass distribution at $z_{max}$
Random measurements: Uniform distribution $p_{rand}$

The model is a mixture of these 3 densities:

$z_{hit} \cdot p_{hit} + z_{rand} \cdot p_{rand} + z_{\max} \cdot p_{\max}$

Issues

Does not explicitly model dynamic objects that cause short readings
Treats sensors as being able to see through walls: ray casting replaced by nearest neighbour function: incapable of determining whether a path to a point is intercepted by an obstacle in the map
Does not account for map uncertainty

These issues can be addressed via extensions to the algorithm.

Links to this note