The Forest Fire Sensor Network simulation required the senors to be in motion. What type of motion should be used? I considered the following:
- Random Brownian Motion.
- Wandering behavior as described by Craig Reynolds, which I first encountered a couple of years ago while reading Daniel Shiffman's Nature of Code
- Predictable orbits resulting from sin and cosine functions
I decided to give the orbits approach a try. Dynamic sensor networks are interesting to visualize because the changing coverage of a region seems highly dependent on the underlying motion of the sensors. A predictable orbit for each sensor allows for the possibility that an evader wishing to remain in the "holes" of the coverage blanket could predict where each hole might form.
We see again the question presented in Robert Ghrist's paper:
The evasion problem for this scenario is whether an unknown evader can navigate through holes in the sensor cover without being detected.