🎮
Pacman

Ghostbusters Inference

250ms
Steps:0
Particle filter with 1000 samples for probabilistic tracking.
Layout
Algo: Particle
Belief mass
0.000
Steps
0
Noisy distance
None
Cells
0
Additional Controls
Inference Settings
How it works
What you see
  • Dark cells are walls. Open cells are traversable.
  • Yellow circle is Pacman; colored sprites are true ghost positions.
  • The red heatmap shows belief over the ghost’s location (probability mass per cell).
Algorithms
  • Exact: full probability table; Bayes update with observation then spread to neighbors.
  • Particle: 1000 samples; weight by sensor (λ=0.2, noise±2), resample, then move (including stay).
Why beliefs differ from sprites
  • Noisy distance fits many symmetric cells; belief can be multi‑modal.
  • Strict sensors (small noise, larger λ) can collapse weights and trigger resets; more particles help.
Real‑world applications
  • Robot localization and tracking (GPS/LiDAR/IMU).
  • Autonomous driving and radar multi‑target tracking.
  • SLAM and AR/VR head tracking.
  • Signal processing and econometrics (state‑space inference).