I have come across the paper that deals with spatial positioning of mobile sensors to optimally detect sound source, or position mobile cellphone towers to maximize the coverage.

The region $Q$ is partitioned into mutually exclusive $n$ Voronoi polytopes $W=\{ W_1,..,W_n \}$. A function $\phi :Q \to \mathbb{R}_+$ assigns probability density that a certain event (here sound source) has happened over $Q$. There are $n$ sensors to be distributed $P=(p_1,..p_n)$ over each Voronoi partition that satisfies the following equation:

$$H(P,W)=\min\sum_{i=1}^n \int_{W_i} f(\| q-p_i\|)\,d \phi(q)$$

The quality of observation at point $q$ from sensor $p_i$ is the distance of the point $q$ from the sensor $f(\| q-p_i\|)$ which makes sense.

Can anyone explain why are we multiplying $(q)$ after $d\phi$ in the objective formula?


1 Answer 1


Note that $\phi$ is a function of $q$ so $d\phi(q)$ is interpreted as w.r.t. $\phi(q)$, which is the same as $d\phi$. This is just shorthand for $\phi'(q)\,dq$.

  • $\begingroup$ Alright, makes so much sense $\endgroup$
    – gfdsal
    May 12, 2020 at 11:08
  • $\begingroup$ I have a small question that is related more to a concept then the notation. Why is the author minimizing the cost function: uncertanity term $f(\| q-p_i\|)$ multiplied by importance term (high density means high importance) $d\phi(q)$. Shouldnt the importance term $d\phi(q)$ require us to maximize the cost function so we place sensors where the density is high? $\endgroup$
    – gfdsal
    May 15, 2020 at 14:55
  • 1
    $\begingroup$ What $H$ aims to do is to minimise the sum of the degradation function $f$ over the partitions; we are not multiplying by the distribution density function but are rather integrating over it like a measure ($Q\to\Bbb R^+$). $\endgroup$
    – TheSimpliFire
    May 15, 2020 at 15:40
  • $\begingroup$ I see. Quite nicely explained! A "measure" is essentially integrating of degradation function over the pdf of the partition. right? $\endgroup$
    – gfdsal
    May 15, 2020 at 16:52

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service and acknowledge you have read our privacy policy.

Not the answer you're looking for? Browse other questions tagged or ask your own question.