Since my last blog post, I’ve been reading the Probabilistic Robotics book. I coded up a Kalman Filter in C++ (https://github.com/taumuon/Kalman-Filter).
It’s been a while since I’d coded up any F#, so I converted a particle filter over from Python. The code is from the online MOOC Artificial Intelligence for Robotics from Udacity.
The idea is to perform robot localisation, by creating a set of particles randomly over the entire state space. On a noisy sensor reading, the probability of that sensor reading having being made for each particle is calculated. Then, all particles undergo a weighted resampling (weighted by the measurement probability).
In this example, the sensor reading is the distance to four beacons at a known position in space. The same algorithm is used in Simultaneous Location and Mapping (SLAM) applications, where the sensor readings could be sonar, radar, visual etc.
The F# code is nice and compact at only 94 lines. The project is available at https://github.com/taumuon/Particle-Filter-FSharp
