1.5-Dimensional Terrain Search

• The 1.5-dimensional terrain search problem is defined by an x-monotone terrain above a horizontal base, establishing a unique framework for online target detection.
• It reduces to a ray searching problem in the upper half-plane, linking classical search theory to geometric art-gallery challenges.
• Algorithmic strategies like the zigzag cow-path method achieve competitive ratios below 9.13, offering near-optimal online search performance.

A 1.5-dimensional terrain search problem refers to the online search for a hidden target along or above a 1.5-dimensional terrain—defined as an $x$-monotone polygonal chain or simple polygon bounded below by a horizontal base segment and above by the (monotone) terrain chain—where detection occurs when a visibility segment from the searcher to the target does not cross below the terrain. The searcher, starting at a known point, traverses a continuous path restricted to lie on or above the terrain, and seeks to optimize the competitive ratio: the worst-case ratio of search path length to the optimal offline cost (shortest possible path from start to detection given full information about the target location). This setting generalizes classical search theory, links to geometric art-gallery problems, and adds nontrivial geometric constraints due to the presence of a terrain barrier.

1.5-Dimensional Terrain Model

A 1.5-dimensional terrain $T$ in $\mathbb{R}^2$ is a simple polygon whose boundary consists of a horizontal base segment $\ell$ (typically on the $x$-axis, endpoints $v_1$ and $v_n$ with $x(v_1)<x(v_2)<\cdots<x(v_n)$) and an upper $x$-monotone polygonal chain connecting $v_1$ through $v_2,\dots,v_{n-1}$ to $v_n$, lying entirely above $\ell$. The internal region $\mathbb{T}$ is the closed polygon bounded by this union.

The searcher's admissible path $\Pi$ is any continuous curve in $\mathbb{R}^2$ that never dips below the terrain, i.e., at all times, $\Pi$ lies on or above $T$.

2. Formal Statement of the Search Problem

Given:

• A 1.5D terrain $T$ as described above.
• A searcher starting at point $s$ on $\partial T$.
• An unknown target $t$ located somewhere on $T$.

Objective:

• Devise an online strategy (path $\Pi$) such that the first point $p_t$ where the line segment $p_t t$ lies entirely above $T$ (i.e., $t$ becomes visible) is reached with minimal competitive ratio:

$$C = \sup_{T, t} \frac{|\Pi_{s}^{p_t}|}{\mathrm{S}_{\mathrm{OPT}}(T,t)},$$

where $\mathrm{S}_{\mathrm{OPT}}(T,t)$ is the length of the shortest (offline) path from $s$ to any point from which $t$ is visible, staying above $T$.

Visibility is defined via visibility rays: $t$ is detected as soon as $\Pi$ crosses a unique ray $r_t$ from some point $u\in T$ toward $s$ such that the segment is never below $T$.

3. Equivalence and Reduction to Ray Searching in the Half-Plane

A crucial structural insight is the reduction of the terrain search problem to a "ray search" in the upper half-plane. Each detection event can be seen as the searcher crossing one of a set $R$ of rays in the half-plane, all passing through some terrain boundary point and directed toward $s$, with the property that the perpendicular from $s$ to the ray's supporting line lies entirely above $T$. Consequently, the difficulty of the original problem is captured by the geometry of the associated rays.

Reduction scheme:

• Any strategy for half-plane ray searching (fixed set of directions upward from $s$) induces a valid terrain search with the same or better competitive ratio, since the offline optimum for terrain is always at least as long as the "perpendicular" (shortest) traverse to the supporting line of the ray.
• Conversely, any adversarial configuration of rays can be realized by constructing a terrain with "spikes" at corresponding positions, forcing the same detection geometry.

This reduction crystallizes the core challenge: ray search in the half-plane, a strict geometric extension of the classic "cow-path" search.

4. Algorithmic Strategies and Competitive Bounds

4.1 Cow-Path-Type Zigzag Algorithm

The principal strategy alternates excursions right and left (with respect to the horizontal axis), with each excursion moving both horizontally and incrementally upward at a prescribed slope, producing a sequence of turn points:

$$p_i = ((-1)^i\,r^i,\,h_i),\quad h_i = 2\,\tan\alpha\sum_{j=0}^{i-1} r^j = 2\,\tan\alpha \frac{r^i-1}{r-1}$$

where $r>1$ is the expansion factor and $\alpha\in (0, \frac\pi2)$ is the angular slope parameter.

The searcher starts at $s$, alternates direction, and walks along these segments, climbing whenever the path would otherwise fall below $T$.

4.2 Competitive Ratio Analysis

Let $\beta \in [\alpha,\pi/2]$ be the direction of a worst-case ray. Two local worst cases dominate:

• Case (a): $\beta=\pi/2$ (vertical rays), where the searcher misses a ray in one excursion and finds it only after the next, producing the largest possible online/offline path length ratio at a fixed turn.
• Case (b): $\beta>\alpha$, a ray tangent at turn $p_{i-1}$.

The closed-form of the asymptotic competitive ratio is

$$C(r,\alpha) = \max_{\beta \in [\alpha,\pi/2]} \frac{[r+1]/\cos \alpha \cdot (r/(r-1) + \sin(\beta-\alpha)/\sin(\beta+\alpha))}{2\cos\beta\sin\alpha(r/(r-1)-1/2) + \cos\alpha\sin\beta}$$

Numerical optimization yields $(r^*,\alpha^*)$ where $C(r^*,\alpha^*) < 9.12725$ and the analytic lower bound is $9.06357$ (Langetepe et al., 18 Dec 2025).

Algorithmic improvements relative to previous estimates tighten the optimality gap to less than $0.064$. The terrain, in effect, cannot force a worse ratio than this pure ray-search setting.

5. Impact of Terrain Geometry and Algorithm Adaptation

A key property is that following the terrain boundary (when upward progress is blocked) never worsens the competitive ratio. Any deviation from the reference path caused by the terrain (switching to "terrain mode" and rejoining at the next upward opportunity) only increases the offline distance needed to see vertical rays, or advances detection time for other rays. Thus, the only truly "hard" instances are terrains mimicking vertical barriers of increasing height—a structural result that sharply restricts adversarial terrains.

6. Key Results and Comparison Table

Bound Type	Value (Exact/Numeric)	Achieved By
Lower bound	9.06357	Adversarial half-circle barrier (Langetepe et al., 18 Dec 2025)
Upper bound (this setting)	<9.12725	Parameter-optimized slope/expansion cow-path (Langetepe et al., 18 Dec 2025)
Previous best upper	9.24663	de Berg et al. (Latin 2024)
Difference (gap)	<0.064	—

This competitive ratio is strictly less than previous best, with the bound achieved by tuning both expansion factor $r$ and slope $\alpha$ and balancing the two boundary cases for $\beta$.

7. Significance and Theoretical Insights

The equivalence of the 1.5D terrain search problem to the ray-search problem in the half-plane reveals a universality class for online geometric search under monotone terrain constraints. The nearly tight bounds—both upper and lower—demonstrate that the primary combinatorial/geometric obstacle is the sequence with which rays at various slopes can be explored, not the detailed geometry of the terrain. The result that exploiting monotonicity and "barrier geometry" can only help (never hurt) the searcher underlines that further improvements to the constant can come only from solving the ray-search instance to optimality.

These results also clarify that within this model, the searcher cannot be forced to do substantially worse than in the planar (non-terrain) case and that the problem reduces, analytically and algorithmically, to the high-precision geometry of rays in the half-plane (Langetepe et al., 18 Dec 2025).