search

jharwell / sierra

Automation framework for the scientific method in AI research
MIT License
18 stars 1 forks source link

feature/162-ta-self-org-measure #162

Closed jharwell closed 4 years ago

jharwell commented 4 years ago

The current self org measure is defined in terms of robot motion, which makes it very homogenous when you compare across task allocation policies. Another measure, which might provide more insight, would be computed from a distance from random task allocation performance (as opposed to random motion). Something like :

Perf policy X with N robots - Perf policy Random with 1 robot * N

Perf policy Random with 1 robot * N

jharwell commented 4 years ago

Some experimentation seemed to show that such a measure would not really provide any additional insight, at least not for my current paper, so tabling for now.