Tight Bounds on Window Size and Time for Single-Agent Graph Exploration under T-Interval Connectivity

Abstract: We study deterministic exploration by a single agent in $T$-interval-connected graphs, a standard model of dynamic networks in which, for every time window of length $T$, the intersection of the graphs within the window is connected. The agent does not know the window size $T$, nor the number of nodes $n$ or edges $m$, and must visit all nodes of the graph. We consider two visibility models, $KT_0$ and $KT_1$, depending on whether the agent can observe the identifiers of neighboring nodes. We investigate two fundamental questions: the minimum window size that guarantees exploration, and the optimal exploration time under sufficiently large window size. For both models, we show that a window size $T = Ω(m)$ is necessary. We also present deterministic algorithms whose required window size is $O(ε(n,m)\cdot m + n \log^2 n)$, where $ε(n,m) = \frac{\ln n}{1 + \ln m - \ln n}$. These bounds are tight for a wide range of $m$, in particular when $m = n^{1+Θ(1)}$. The same algorithms also yield optimal or near-optimal exploration time: we prove lower bounds of $Ω((m - n + 1)n)$ in the $KT_0$ model and $Ω(m)$ in the $KT_1$ model, and show that our algorithms match these bounds up to a polylogarithmic factor, while being fully time-optimal when $m = n^{1+Θ(1)}$. This yields tight bounds when parameterized solely by $n$: $Θ(n^3)$ for $KT_0$ and $Θ(n^2)$ for $KT_1$.

• The paper establishes tight bounds on the minimum window size (Θ(m)) and exploration time for both KT0 and KT1 visibility models.
• It introduces deterministic, label-oblivious algorithms using adversarial gadgets and dynamic map construction to tackle exploration in T-interval-connected graphs.
• The study underscores that even slight improvements in local visibility dramatically reduce exploration time, revealing key trade-offs in dynamic network exploration.

Tight Bounds for Single-Agent Exploration in $T$-Interval-Connected Graphs

Introduction

This paper tackles deterministic single-agent exploration in $T$-interval-connected dynamic graphs, a standard model for dynamic networks where the intersection of all edge-sets in any window of $T$ consecutive time-steps is connected. The agent possesses no global knowledge concerning the window size $T$, the number of vertices $n$, or the number of edges $m$. The work systematically establishes tight bounds on the minimum window size required to guarantee exploration and optimal exploration time under two visibility models: $KT_0$ (minimal port-only visibility, no local neighbor IDs) and $KT_1$ (one-hop neighbor IDs are visible).

Model and Problem Formalization

Let $G=(V,E)$ denote the underlying network. The dynamic nature is encoded by temporal edge-sets $E(t) \subseteq E$. $T$0-interval-connectivity asserts that for any interval $T$1, the intersection graph $T$2 is connected. The agent must visit all vertices, starting from an arbitrary position, with no a priori knowledge of $T$3, $T$4, or $T$5.

• $T$6: the agent sees only port numbers for available outgoing edges at its current location.
• $T$7: the agent additionally observes the identifiers of neighboring nodes.

The core combinatorial quantities studied are:

• $T$8: minimum window size $T$9 such that, with visibility model $T$0, there exists a deterministic exploration algorithm for all $T$1-interval-connected graphs with $T$2 nodes and $T$3 edges.
• Exploration time: the total number of agent moves needed, given sufficiently large $T$4.

Main Results

Minimum Window Size: Tight Bounds

The authors establish that in both visibility settings, any deterministic exploration algorithm requires and suffices with a window $T$5 for all large enough $T$6:

• Lower Bound: $T$7 is proven using gadgets that adversarially confine the agent for $T$8 steps unless the window is at least $T$9 (see Lemma 11 and Theorem~4; examples depicted below).

  Figure 1: $T$0 (left) and $T$1 (right), which serve as adversarial gadgets forcing $T$2 delay before the agent can reach a "gate" node.

  

  Figure 2: The underlying graph for the lower bound construction in Theorem~4, demonstrating the trap architecture.

• Upper Bound: The paper presents deterministic label-oblivious algorithms for both $T$3 and $T$4 requiring window size $T$5, where $T$6. The upper and lower bounds match up to polylogarithmic factors and are asymptotically tight when $T$7.
• When parameterized solely by $T$8, these results yield $T$9.

Exploration Time: Tight Bounds

Given sufficiently large $n$0, optimal exploration time matches lower bounds:

• $n$1: Time is $n$2, matching the minimum window up to small factors. The lower bound is $n$3.
• $n$4: Time is $n$5; the lower bound is $n$6.

  Figure 3: The underlying graph for the $n$7 lower bound in the $n$8 model, showing how redundant exploration over unknown port assignments induces worst-case behavior.

Further, for $n$9 the time and window size are matched. For $m$0, exploration incurs an extra multiplicative factor of $m$1 due to the absence of neighbor identity information, an unavoidable overhead supported by lower-bound constructions.

Algorithmic Techniques and Proof Insights

• Adversarial Gadgets: The $m$2 family (complete graph with one missing edge) allows the adversary to confine the agent among "trap" vertices, utilizing minimal temporal edge deletions, while preserving $m$3-interval-connectivity. This strategy leverages combinatorial properties of the gadget such that, unless the window is wide enough, exploration of all nodes is infeasible.
• Greedy Exploration with Map Construction (for $m$4): The presented algorithm builds an evolving map, dynamically eliminates unavailable edges, and greedily searches for the closest unvisited node by shortest path, guaranteeing progress on the always-present spanning tree (shown to exist by a window-size argument).
• Redundant Traversals in $m$5: In the absence of neighbor IDs, explicit resetting, and detection of unused ports is critical. The algorithm must repeatedly attempt all possible unexplored ports, which, given adversarial port assignments, result in $m$6 extra steps per edge, thus inflating asymptotic exploration time.

Numerical & Theoretical Highlights

• All upper and lower bounds are tight for $m$7 and for $m$8 as the main parameter (i.e., clique-like graphs).
• The precise gap between upper and lower bounds for general $m$9 is confined to an $KT_0$0 multiplicative factor.
• The algorithms make no use of global parameters, such as $KT_0$1, $KT_0$2, or $KT_0$3, and operate under adversarial, strongly dynamic conditions.

Implications and Future Work

The results precisely characterize deterministic single-agent exploration in highly dynamic environments, establishing that the minimal visibility jump from $KT_0$4 to $KT_0$5 yields a linear-to-quadratic reduction in exploration time, and that the bottleneck is inherently the topological edge count $KT_0$6 under adversarial schedules. Practically, this means that efficient deterministic exploration is feasible only if interval connectivity is, on average, commensurate with the edge set size, and that small changes in local visibility yield substantial algorithmic gains.

An open line is the potential of randomization: the authors note that randomization could potentially break through the linear-in-$KT_0$7 window-size barrier or reduce the quadratic time in $KT_0$8. Other future directions include precisely quantifying the polylogarithmic gap for irregular or sparse graphs and analyzing the trade-off between memory complexity and exploration efficiency.

Conclusion

The paper settles the deterministic single-agent interval-connectivity exploration problem for arbitrary $KT_0$9 and $KT_1$0, deriving tight window size and time bounds for both weak and strong local observation models. The lower bounds are established through robust adversarial constructions, while the upper bounds leverage efficient, memoryless (apart from map construction) exploration strategies. This work provides clarity on the interplay between structural dynamics, visibility, and exploration complexity in temporal graphs, establishing a foundational reference point for future algorithms and lower bound studies in the domain of mobile agents and network exploration.