Optimal Sensor Placement for Intruder Detection

We consider the centralized detection of an intruder, whose location is modeled as uniform across a specified set of points, using an optimally placed team of sensors. These sensors make conditionally independent observations. The local detectors at the sensors are also assumed to be identical, with detection probability $(P_{{D}})$ and false alarm probability $(P{{F}})$. We formulate the problem as an N-ary hypothesis testing problem, jointly optimizing the sensor placement and detection policies at the fusion center. We prove that uniform sensor placement is never strictly optimal when the number of sensors $(M)$ equals the number of placement points $(N)$. We prove that for $N{2} > N_{1} > M$, where $N_{1},N_{2}$ are number of placement points, the framework utilizing $M$ sensors and $N_{1}$ placement points has the same optimal placement structure as the one utilizing $M$ sensors and $N_{2}$ placement points. For $M\leq 5$ and for fixed $P_{{D}}$, increasing $P{{F}}$ leads to optimal placements that are higher in the majorization-based placement scale. Similarly for $M\leq 5$ and for fixed $P{{F}}$, increasing $P{_{D}}$ leads to optimal placements that are higher in the majorization-based placement scale. For $M>5$, this result does not necessarily hold and we provide a simple counterexample. It is conjectured that the set of optimal placements for a given $(M,N)$ can always be placed on a majorization-based placement scale.

💡 Research Summary

This paper investigates the joint design of sensor placement and fusion‑center decision rules for centralized detection of an intruder whose location is uniformly distributed over a finite set of N points. A team of M identical binary sensors is deployed; each sensor reports a 0 (no intruder) or 1 (intruder present) observation that is conditionally independent given the true intruder location. The sensors are characterized by a detection probability (P_D) and a false‑alarm probability (P_F), both assumed known a priori. The performance metric is the overall probability of error (P_e) at the fusion center, which employs a Bayesian N‑ary hypothesis test with uniform priors, leading to a maximum‑a‑posteriori (MAP) decision rule.

The authors formulate the problem as an optimization over the integer‑partition set (\Lambda_M) of all possible placement vectors (\mathbf{v} = (v_1,\dots,v_N)) where (v_i) denotes how many sensors are placed at point i and (\sum_i v_i = M). Because the size of (\Lambda_M) grows as the partition function (f(M)) (approximately (\frac{1}{4\sqrt{3}M}e^{\pi\sqrt{2M/3}})), exhaustive search is infeasible for moderate M. To obtain structural insight, the paper leverages majorization theory. A placement (\mathbf{v}^{(1)}) majorizes (\mathbf{v}^{(2)}) ((\mathbf{v}^{(1)}\succ\mathbf{v}^{(2)})) if the sorted partial sums of (\mathbf{v}^{(1)}) dominate those of (\mathbf{v}^{(2)}). This defines a partial order—called the majorization‑based placement scale—on the set of feasible placements.

Key theoretical contributions are:

1. Uniform placement is never strictly optimal when M = N. Using induction, the authors first prove the case M = N = 2 by explicit calculation of (P_e) for the uniform placement (1,1) and the concentrated placement (2,0). They show that the latter always yields a smaller or equal error probability for any admissible ((P_D,P_F)). The induction step extends this result to arbitrary k, establishing that the uniform vector ((1,\dots,1)) cannot be strictly optimal for any ((P_D,P_F)).

2. Invariance of optimal structure with respect to the number of placement points. For any two numbers of locations (N_2 > N_1 > M), the optimal placement pattern (ignoring zero‑sensor locations) is identical. In other words, adding extra “empty” locations does not alter which sensors are grouped together in the optimal solution.

3. Monotonicity of optimal placements on the majorization scale for small sensor counts. When the number of sensors satisfies (M \le 5):

   • Fixing (P_D) and increasing (P_F) moves the optimal placement upward on the majorization scale (i.e., toward more concentrated configurations).
   • Fixing (P_F) and increasing (P_D) has the same effect. This aligns with intuition: poorer sensor reliability (higher false alarms) or better detection capability (higher (P_D)) both favor concentrating sensors on fewer points to reduce overall error.

4. Counter‑example for larger sensor counts. For (M > 5) the monotonicity property no longer holds universally. The authors present a simple example with (M = 6) where raising (P_F) does not shift the optimal placement upward, demonstrating that the majorization‑based ordering can be broken when many sensors are available.

5. Conjecture on universal majorization ordering. Based on the proven cases, the paper conjectures that for any ((M,N)) pair the set of optimal placements can be arranged on a majorization‑based scale, i.e., there exists a total ordering consistent with majorization that captures all optimal solutions.

Simulation studies corroborate the analytical findings. Exhaustive enumeration for small M and N confirms that the optimal placements follow the predicted majorization ordering when (M \le 5). For larger M, the simulations reveal the occasional violation of the monotonicity rule, matching the theoretical counter‑example.

In conclusion, the work provides a rigorous framework that links sensor reliability parameters to optimal spatial deployment via majorization theory. It offers practical design rules for systems with up to five sensors—where one can simply select the most “majorized” placement consistent with the given (P_D) and (P_F). The paper also outlines future research directions, including extending the analysis to non‑uniform priors, multiple intruders, correlated sensor observations, and proving (or disproving) the conjectured universal majorization ordering.