Spike landmarks

Source: SLAM for Dummies

• Uses extrema to find landmarks
• Find values in the range of a laser scan, where two values differ by more than a certain amount (e.g. 0.5 m)
  • This finds big changes in the laser scan
• Alternatively, using three values next to each other: $A, B, C$
  • $(A - B) + (C - B)$ yields a value
  • Better for finding spikes as it finds actual spikes
• Rely on the landscape changing a lot between two laser beams
  • Algo will fail in smooth environments
• Suitable for indoor environments, however is not robust against envs w/ people
  • people are picked up as spikes as theoretically they are good landmarks (just not stationary!)