Pinning Balloons with Perfect Angles and Optimal Area

We study the problem of arranging a set of $n$ disks with prescribed radii on $n$ rays emanating from the origin such that two neighboring rays are separated by an angle of $2\pi/n$. The center of the disks have to lie on the rays, and no two disk centers are allowed to lie on the same ray. We require that the disks have disjoint interiors, and that for every ray the segment between the origin and the boundary of its associated disk avoids the interior of the disks. Let $\r$ be the sum of the disk radii. We introduce a greedy strategy that constructs such a disk arrangement that can be covered with a disk centered at the origin whose radius is at most $2\r$, which is best possible. The greedy strategy needs O(n) arithmetic operations. As an application of our result we present an algorithm for embedding unordered trees with straight lines and perfect angular resolution such that it can be covered with a disk of radius $n^{3.0367}$, while having no edge of length smaller than 1. The tree drawing algorithm is an enhancement of a recent result by Duncan et al. [Symp. of Graph Drawing, 2010] that exploits the heavy-edge tree decomposition technique to construct a drawing of the tree that can be covered with a disk of radius $2 n^4$.

💡 Research Summary

The paper tackles two closely related problems: (1) a geometric packing problem that we call the “balloon layout problem,” and (2) an application of this packing to the drawing of unordered trees with perfect angular resolution.

In the balloon layout problem we are given n disks (balloons) with prescribed radii r₁,…,rₙ and n rays emanating from the origin, the rays being equally spaced by an angle 2π/n. Each balloon must be placed with its centre on a distinct ray, the interiors of the balloons must be pairwise disjoint, and the segment from the origin to the boundary of a balloon must not intersect any other balloon. Let r = Σ rᵢ be the total sum of the radii. The authors present a greedy algorithm that constructs a placement satisfying all constraints and guarantees that the whole configuration can be covered by a single disk centred at the origin whose radius is at most 2r. This bound is tight: when n = 1 the optimal covering radius is exactly 2r₁.

The algorithm works in O(n) arithmetic operations. It processes the balloons in non‑decreasing order of radius. The placement proceeds in rounds; in each round roughly half of the still‑unused rays are selected (every second ray in cyclic order). The selected rays are spaced by an angle αᵢ = 2^{i+1}π/n, which doubles each round. A “safe disk” SDᵢ is maintained – the smallest disk centred at the origin that already covers all balloons placed in earlier rounds. In the current round each balloon is first attempted to be placed so that it touches the boundary of SDᵢ while staying on its assigned ray. If this would violate the angular wedge defined by the neighbouring rays, the balloon is moved outward along its ray until it touches the wedge boundary; this is called a wedge situation. Lemma 1 shows that in a wedge situation the balloon is contained in a disk of radius α(ϕ)·r where α(ϕ)=1+sin(ϕ/2)/sin(ϕ/2) and ϕ is the wedge angle. The algorithm therefore computes the exact distance from the origin for each balloon using either the “contact” distance (touching SDᵢ) or the wedge distance, whichever is larger.

When the number of remaining rays becomes odd, the algorithm uses a “well‑separated” selection (Algorithm 1) that guarantees the two smallest angles among the remaining rays are at least 2π/3, ensuring that the final two balloons can be placed: the largest balloon touches the safe disk, and the second‑largest is placed inside a wedge of angle π/3. Lemma 3 shows that either the last layer has width 2·rₙ or the whole configuration can already be covered by a disk of radius 2·r, establishing the 2r bound.

Having solved the geometric sub‑problem, the authors apply it to tree drawing. Each vertex of an unordered tree is represented by a balloon whose radius encodes a weight (or simply 1 for uniform size). The children of a node are drawn as sub‑trees, each already laid out inside its own balloon. The greedy balloon placement is then used to arrange these child balloons around the parent’s ray, preserving perfect angular resolution: incident edges at any vertex are separated by at least 2π/degree(v). Because the greedy algorithm guarantees a covering radius of at most 2r for the set of child balloons, the recursive construction yields a drawing that can be covered by a disk of radius O(n^{3.0367}) while ensuring that every edge has length at least 1. This improves on the previous best known bound of O(n⁴) by Duncan et al. (2010).

The paper also discusses extensions for the case where n is not a power of two, handling odd numbers of rays, and proving that the “well‑separated” property is preserved through the recursive splitting. The analysis includes several lemmas that bound the growth of the covering radius, showing that the worst case occurs when the largest balloons dominate the later layers and the smaller ones are set to zero (Lemma 5).

In the context of related work, the authors compare their results with bubble drawings, ringed circular layouts, and orthogonal straight‑line drawings of ternary trees, noting that those either sacrifice angular resolution or require super‑polynomial area. Their contribution is the first to achieve both perfect angular resolution and near‑optimal area (within a factor of 2 of the trivial lower bound) for unordered trees, and to provide a linear‑time algorithm for the underlying geometric packing problem.

Overall, the paper delivers a clean, optimal greedy strategy for arranging disks with prescribed radii on equally spaced rays, proves that the resulting layout is area‑optimal up to a factor of 2, and leverages this result to obtain significantly improved tree drawing algorithms with perfect angular resolution and polynomial area.