Cops-and-Robber Entanglement Game

• The cops-and-robber entanglement game is a pursuit-evasion process on finite graphs where cops can only teleport to the robber's current vertex, emphasizing reactive over pre-positioning strategies.
• Key results show that the entanglement number is bounded below by the graph’s minimum degree, with proofs that leverage probabilistic methods and combinatorial arguments even for sparse graphs.
• The study employs techniques like cutting arguments, peeling methods, and chasing strategies to reveal surprising lower bounds and open questions regarding optimal cop deployment on various graph classes.

The cops-and-robber entanglement game is a pursuit-evasion process on finite graphs that restricts the placement of pursuing agents ("cops") relative to a moving adversary ("robber"), revealing deep combinatorial structure and surprising lower bounds even for sparse and low-degree graphs. Unlike the classical cops-and-robbers game, where cops may occupy arbitrary vertices, in the entanglement variant the cops may only "fly" one at a time to the robber's current position and cannot otherwise pre-position themselves anywhere on the graph except by following the robber. This change in protocol fundamentally alters both optimal strategies and required resources. The concept was systematically studied in "Even flying cops should think ahead" (Martinsson, Meier, Schnider & Steger, 2018) (Martinsson et al., 2018), which introduced the formal definition, key theorems, combinatorial techniques, and several open questions.

1. Formal Definition of the Entanglement Game

Let $G=(V,E)$ be a finite (undirected or directed) graph and fix an integer $k \ge 0$. The entanglement game with $k$ cops and one robber is played in discrete rounds:

• Setup: The robber first chooses a vertex $r_0 \in V$. All $k$ cops start off-graph.
• Each round $t = 0, 1, 2, \dots$:

1. The cops either do nothing, or select exactly one cop to "fly" (teleport, ignoring edges) to the robber's current position $r_t$. The other cops remain stationary. 2. The robber must move from $r_t$ to a neighbor $r_{t+1}$ of $r_t$ not occupied by any cop. If no such neighbor exists, the cops win; otherwise, play proceeds.

A configuration can be represented as $k \ge 0$0, where $k \ge 0$1 is the robber's position and $k \ge 0$2 the set of vertices currently occupied by cops (with at most one cop per vertex). The entanglement number of $k \ge 0$3, denoted $k \ge 0$4, is the smallest $k \ge 0$5 such that $k \ge 0$6 cops have a winning strategy.

2. Contrasts with the Classical Cops-and-Robbers Game

Unlike the classical game—where all cops move along edges simultaneously and may preemptively block vertices—the entanglement game imposes several restrictions:

• Cops may only teleport directly to the robber's current vertex, and only one cop may be moved per round.
• Cops cannot occupy arbitrary vertices, nor can they pre-position themselves on cut-vertices or in corridors unless they have just followed the robber.
• Cornering the robber is generally impossible by pre-placement; instead, cops must reactively attempt to contain the robber's movement.

On trees, two cops suffice in the classical game, and the same remains true for entanglement, but the nature of the proof differs, as entangled cops cannot surround tree branches preemptively (Martinsson et al., 2018).

3. Core Theorems and Key Results

Minimum-degree lower bound:

For any graph $k \ge 0$7, the robber can always evade fewer than $k \ge 0$8 cops, where $k \ge 0$9; thus

$k$0

Sparse random graphs:

For every $k$1 there exists $k$2 so that, for $k$3, with high probability,

$k$4

where $k$5 denotes the standard Erdős–Rényi random graph. The proof uses a cut-counting argument to ensure that any sufficiently large cop-free set contains a large connected component, allowing the robber to maintain safe escaping options (Martinsson et al., 2018).

3-regular graphs with large entanglement:

If $k$6 are three random perfect matchings on $k$7 (even) vertices, then with high probability

$k$8

for some constant $k$9. Thus, there exist infinitely many simple 3-regular graphs with linear entanglement number.

Upper and lower bounds summary:

Graph Class	Lower Bound	Upper Bound	Notes
Any finite $r_0 \in V$0	$r_0 \in V$1	$r_0 \in V$2	$r_0 \in V$3
Tree	$r_0 \in V$4	$r_0 \in V$5	Holds by alternative arguments
2-regular (disjoint cycles)	$r_0 \in V$6	$r_0 \in V$7
3-regular	$r_0 \in V$8	$r_0 \in V$9	Large gap established by Martinsson et al. (Martinsson et al., 2018)

Explicit small entanglement for a 3-regular family:

For the 3-regular graph $k$0 obtained from a cycle with added antipodal diagonals (on $k$1 vertices), $k$2. The proof constructs a chase-and-block strategy using pairs of cops to successively block diagonals, reducing the robber's options.

Directed graphs of total degree 3:

By "blowing up" each vertex of a hard 3-regular undirected graph into a directed 6-cycle and appropriately redirecting edges, there exist directed graphs on $k$3 vertices of maximum total degree 3 and entanglement number $k$4.

4. Techniques and Proof Methodologies

Key methodologies include:

• Probabilistic method and large-deviation bounds: Chernoff and entropy inequalities are employed to establish expansion properties in random graphs and unions of perfect matchings.
• Combinatorial partition lemma: If a graph contains no component larger than half its vertices, its vertex set can be 3-colored so that no edges cross color classes—a property crucial for analyzing cop-free safe sets.
• Stirling’s approximation and union bounds: Used to estimate the probability that random perfect matchings avoid certain structures, informing robustness properties.
• Peeling (kernelization) arguments: Applied to 3-regular graphs to systematically reduce the graph to bounded-width components, each manageable by simple cop strategies.
• Chasing and blocking: Constructing explicit cop strategies, particularly in low-degree or reduced graphs, by alternately pursuing and permanently blocking paths narrowed by the robber's forced movement.

5. General Bounds and Special Graph Classes

The relationship between structural parameters and entanglement is highly nontrivial. In particular:

• The minimum degree provides a universal lower bound, but even for graphs of maximum degree 3, entanglement can be linear in the number of vertices.
• For complete graphs, maximum entanglement is achieved: $k$5.
• For trees, $k$6, and for 2-regular graphs, $k$7.
• For 3-regular graphs, a gap persists between the linear lower bound $k$8 and the upper bound $k$9, with explicit examples demonstrating both behaviors (Martinsson et al., 2018).

6. Open Questions and Research Directions

Several foundational questions remain unresolved:

• Determining the sharp constant $t = 0, 1, 2, \dots$0 such that, with high probability, the entanglement number of a union of three random perfect matchings on $t = 0, 1, 2, \dots$1 vertices is approximately $t = 0, 1, 2, \dots$2.
• Closing the gap between the upper bound $t = 0, 1, 2, \dots$3 and the linear lower bound for 3-regular graphs.
• Characterizing entanglement for other sparse families, such as planar, bounded genus, or minor-excluded graphs, in terms of structural parameters like treewidth or size of a feedback vertex set.
• Complexity-theoretic questions: while deciding $t = 0, 1, 2, \dots$4 is polynomial-time solvable for certain classes (e.g., directed graphs with bounded entanglement), the computational complexity for general graphs remains open (Martinsson et al., 2018).

The entanglement game thus presents a rich landscape of combinatorial and probabilistic phenomena, exposing the limitations of classical pursuit strategies and raising deep questions regarding graph structure and algorithmic complexity. Even in graphs with sparse connectivity, cycles that are sufficiently "entangled" can necessitate a linear number of cops, sharply contrasting with classical results and demanding new analytical tools.