Sheung-Hung Poon

Fáry’s Theorem for 1-Planar Graphs

A plane graph is a graph embedded in a plane without edge crossings. Fary’s theorem states that every plane graph can be drawn as a straight-line drawing, preserving the embedding of the plane graph. In this paper, we extend Fary’s theorem to a class of non-planar graphs. More specifically, we study the problem of drawing 1-plane graphs with straight-line edges. A 1-plane graph is a graph embedded in a plane with at most one crossing per edge. We give a characterisation of those 1-plane graphs that admit a straight-line drawing. The proof of the characterisation consists of a linear time testing algorithm and a drawing algorithm. Further, we show that there are 1-plane graphs for which every straight-line drawing has exponential area. To the best of our knowledge, this is the first result to extend Fary’s theorem to non-planar graphs.

Optimizing active ranges for consistent dynamic map labeling

Map labeling encounters unique issues in the context of dynamic maps with continuous zooming and panning-an application with increasing practical importance. In consistent dynamic map labeling, distracting behavior such as popping and jumping is avoided. In the model for consistent dynamic labeling that we use, a label becomes a 3d-solid, with scale as the third dimension. Each solid can be truncated to a single scale interval, called its active range, corresponding to the scales at which the label will be selected. The active range optimization (ARO) problem is to select active ranges so that no two truncated solids overlap and the sum of the heights of the active ranges is maximized. The simple ARO problem is a variant in which the active ranges are restricted so that a label is never deselected when zooming in. We investigate both the general and simple variants, for 1d- as well as 2d-maps. The 1d-problem can be seen as a scheduling problem with geometric constraints, and is also closely related to geometric maximum independent set problems. Different label shapes define different ARO variants. We show that 2d-ARO and general 1d-ARO are NP-complete, even for quite simple shapes. We solve simple 1d-ARO optimally with dynamic programming, and present a toolbox of algorithms that yield constant-factor approximations for a number of 1d- and 2d-variants.

Curve reconstruction from noisy samples

We present an algorithm to reconstruct a collection of disjoint smooth closed curves from n noisy samples. Our noise model assumes that the samples are obtained by first drawing points on the curves according to a locally uniform distribution followed by a uniform perturbation of each point in the normal direction with a magnitude smaller than the minimum local feature size. The reconstruction is faithful with a probability that approaches 1 as n increases.We expect that our approach can lead to provable algorithms under less restrictive noise models and for handling non-smooth features.

Labeling Points with Weights

Abstract Annotating maps, graphs, and diagrams with pieces of text is an important step in information visualization that is usually referred to as label placement. We define nine label-placement models for labeling points with axis-parallel rectangles given a weight for each point. There are two groups: fixed-position models and slider models. We aim to maximize the weight sum of those points that receive a label. We first compare our models by giving bounds for the ratios between the weights of maximum-weight labelings in different models. Then we present algorithms for labeling n points with unit-height rectangles. We show how an O(n\log n)-time factor-2 approximation algorithm and a PTAS for fixed-position models can be extended to handle the weighted case. Our main contribution is the first algorithm for weighted sliding labels. Its approximation factor is (2+\varepsilon), it runs in O(n2/\varepsilon) time and uses O(n/\varepsilon) space. We show that other than for fixed-position models even the projection to one dimension remains NP-hard. For slider models we also investigate some special cases, namely (a) the number of different point weights is bounded, (b) all labels are unit squares, and (c) the ratio between maximum and minimum label height is bounded.

Hierarchy of surface models and irreducible triangulations

Given a triangulated closed surface, the problem of constructing a hierarchy of surface models of decreasing level of detail has attracted much attention in computer graphics. A hierarchy provides view-dependent refinement and facilitates the computation of parameterization. For a triangulated closed surface of n vertices and genus g, we prove that there is a constant c > 0 such that if n > c ċ g, a greedy strategy can identify Θ(n) topology-preserving edge contractions that do not interfere with each other. Further, each of them affects only a constant number of triangles. Repeatedly identifying and contracting such edges produces a topology-preserving hierarchy of O(n + g2) size and O(logn + g) depth. Although several implementations exist for constructing hierarchies, our work is the first to show that a greedy algorithm can efficiently compute a hierarchy of provably small size and low depth. When no contractible edge exists, the triangulation is irreducible. Nakamoto and Ota showed that any irreducible triangulation of an orientable 2-manifold has at most max{342g - 72, 4} vertices. Using our proof techniques we obtain a new bound of max{240g, 4}.

Kinetic collision detection for convex fat objects

We design compact and responsive kinetic data structures for detecting collisions between n convex fat objects in 3-dimensional space that can have arbitrary sizes. Our main results are: (i) If the objects are 3-dimensional balls that roll on a plane, then we can detect collisions with a KDS of size O(nlogn) that can handle events in O (logn) time. This structure processes O(n2) events in the worst case, assuming that the objects follow constant-degree algebraic trajectories. (ii) If the objects are convex fat 3-dimensional objects of constant complexity that are free-flying in R3, then we can detect collisions with a KDS of O(nlog6n) size that can handle events in O(log6n) time. This structure processes O(n2) events in the worst case, assuming that the objects follow constant-degree algebraic trajectories. If the objects have similar sizes then the size of the KDS becomes O(n) and events can be handled in O(1) time.

Towards a hybrid approach to SoC estimation for a smart Battery Management System (BMS) and battery supported Cyber-Physical Systems (CPS)

One of the most important and indispensable parameters of a Battery Management System (BMS) is to accurately estimate the State of Charge (SoC) of battery. Precise estimation of SoC can prevent battery from damage or premature aging by avoiding over charge or discharge. Due to the limited capacity of a battery, advanced methods must be used to estimate precisely the SoC in order to keep battery safely being charged and discharged at a suitable level and to prolong its life cycle. We review several existing effective approaches such as Coulomb counting, Open Circuit Voltage (OCV) and Kalman Filter method for performing the SoC estimation. Then we investigate both Artificial Intelligence (AI) approach and Formal Methods (FM) approach that can be efficiently used to precisely determine the SoC estimation for the smart battery management system as presented in [1]. By using presented approach, a more accurate SoC measurement can be obtained for the smart battery management system and battery supported Cyber-Physical Systems (CPS).

On straightening low-diameter unit trees

A polygonal chain is a sequence of consecutively joined edges embedded in space. A k-chain is a chain of k edges. A polygonal tree is a set of edges joined into a tree structure embedded in space. A unit tree is a tree with only edges of unit length. A chain or a tree is simple if non-adjacent edges do not intersect. We consider the problem about the reconfiguration of a simple chain or tree through a series of continuous motions such that the lengths of all tree edges are preserved and no edge crossings are allowed. A chain or tree can be straightened if all its edges can be aligned along a common straight line such that each edge points “away” from a designed leaf node. Otherwise it is called locked. Graph reconfiguration problems have wide applications in contexts including robotics, molecular conformation, rigidity and knot theory. The motivation for us to study unit trees is that for instance, the bonding-lengths in molecules are often similar, as are the segments of robot arms.

On rectilinear drawing of graphs

A rectilinear drawing is an orthogonal grid drawing without bends, possibly with edge crossings, without any overlapping between edges, between vertices, or between edges and vertices. Rectilinear drawings without edge crossings (planar rectilinear drawings) have been extensively investigated in graph drawing. Testing rectilinear planarity of a graph is NP-complete [10]. Restricted cases of the planar rectilinear drawing problem, sometimes called the “no-bend orthogonal drawing problem”, have been well studied (see, for example,[13],[14],[15] ). In this paper, we study the problem of general non-planar rectilinear drawing; this problem has not received as much attention as the planar case. We consider a number of restricted classes of graphs and obtain a polynomial time algorithm, NP-hardness results, an FPT algorithm, and some bounds. We define a structure called a “4-cycle block”. We give a linear time algorithm to test whether a graph that consists of a single 4-cycle block has a rectilinear drawing, and draw it if such a drawing exists. We show that the problem is NP-hard for the graphs that consist of 4-cycle blocks connected by single edges, as well as the case where each vertex has degree 2 or 4. We present a linear time fixed-parameter tractable algorithm to test whether a degree-4 graph has a rectilinear drawing, where the parameter is the number of degree-3 and degree-4 vertices of the graph. We also present a lower bound on the area of rectilinear drawings, and a upper bound on the number of edges.

Boundary Labeling with Flexible Label Positions

Boundary labeling connects each point site in a rectangular map to a label on the sides of the map by a leader, which may be a straight-line segment or a polyline. In the conventional setting, the labels along a side of the map form a single stack of labels in which labels are placed consecutively one by one in a sequence, and the two end sides of a label stack must respect the sides of the map. However, such a setting may be in conflict with generation of a better boundary labeling, measured by the total leader length or the number of bends of leaders. As a result, this paper relaxes this setting to propose the boundary labeling with flexible label positions, in which labels are allowed to be placed at any non-overlapping location along the sides of the map so that they do not necessarily form only one single stack, and the two end sides of label stacks do not need to respect the sides of the map. In this scenario, we investigate the total leader length minimization problem and the total bend minimization problem under several variants, which are parameterized by the number of sides to which labels are attached, their label size, port types, and leader types. It turns out that almost all of the total leader length minimization problems using nonuniform-size labels are NP-complete except for one case, while the others can be solved in polynomial time.