The rapid increase in high-rise building construction has intensified the need for efficient elevator system operations. This article addresses the elevator dispatching problem in elevator group control systems. We formulate the problem as a Semi-Markov Decision Process (SMDP), defining the state representation, action space, and reward function. A two-phase model is then introduced, integrating imitation learning and deep reinforcement learning techniques to derive the optimal elevator dispatching policy from the formulated SMDP. In the first phase, a policy network is pre-trained by estimating the time required for elevator cars to pick up assigned hall requests. In the second phase, the pre-trained policy network is further optimized using Proximal Policy Optimization (PPO), a well-known policy-based deep reinforcement learning method. Additionally, we propose a novel update interval, termed the “direct-effect” interval, which improves policy training during the reinforcement learning phase. Notably, this direct-effect interval concept has potential applicability to other multi-resource scheduling problems. Empirical experiments demonstrate the advantages of incorporating imitation learning before reinforcement learning, as well as the effectiveness of employing the direct-effect update interval during the reinforcement learning phase. Furthermore, the proposed model outperforms various benchmark rules in terms of average waiting time and the distribution of long waiting times, as validated across four traffic patterns.

Most previous studies on multi-agent reinforcement learning focus on deriving decentralized and cooperative policies to maximize a common reward and rarely consider the transferability of trained policies to new tasks. This prevents such policies from being applied to more complex multi-agent tasks. To resolve these limitations, we propose a model that conducts both representation learning for multiple agents using hierarchical graph attention network and policy learning using multi-agent actor-critic. The hierarchical graph attention network is specially designed to model the hierarchical relationships among multiple agents that either cooperate or compete with each other to derive more advanced strategic policies. Two attention networks, the inter-agent and inter-group attention layers, are used to effectively model individual and group level interactions, respectively. The two attention networks have been proven to facilitate the transfer of learned policies to new tasks with different agent compositions and allow one to interpret the learned strategies. Empirically, we demonstrate that the proposed model outperforms existing methods in several mixed cooperative and competitive tasks.

Equipment usually consists of many components arranged in hierarchical structure. In order to achieve efficient maintenance strategy, the system hierarchy should be taken into account. In this paper, we first give a nomenclature to describe a system composed of multiple non-identical components in a hierarchical structure, the system for an age-based and an opportunistic preventive maintenance strategies is modeled by using a Markov Decision Process (MDP). Then, near-optimal policies are found through the SARSA(λ) algorithm from Reinforcement Learning (RL), where the expected discounted cost is minimized. Simulation experiments to compare near-optimal policies obtained by SARSA(λ) are performed for both strategies with corrective maintenance and with age-based preventive maintenance policy obtained from renewal reward theory. We show that the proposed opportunistic preventive maintenance outperforms other strategies.

In this paper, we present a mitered offsetting method of a triangular mesh. Though our main target application is machining tool path generation, it can also be applied to shelling/hollowing of solid objects, collision avoidance in robot path planning, and so on. Previous literature on mesh offsetting mostly suggest inserting a portion of a cylinder (or a ball) in order to fill the gap between offset faces adjacent to a sharp edge (or a sharp vertex, respectively). The gap filling elements (cylinders or balls) are approximated by a number of small triangles depending on the offset error tolerance. Those small gap filling triangles not only increase tool path computation time, but also cause harmful effect in the accuracy of the machined result around the sharp edges. In this research, we try to reduce the number of gap filling triangles while meeting the given tolerance by introducing the concept of mitered offset, which is popularly used in 2D profile machining practice. We borrowed and modified the notion of quadric error metric (QEM) from the mesh simplification area. A modified version of QEM is used for robust computation of the offset vertex position which minimizes the sum of squared distance error from the faces around the original mesh vertex. If the error is within tolerance, the offset vertex is accepted. Otherwise, the offset vertex is split repeatedly until the error is acceptable. Vertex split occurs at the sharp features. A rigorous foundation is given to the mitered offset of 3D mesh with sharp features as well as smooth regions. The experimental results indicate that only a small number of triangles are added in offset mesh.

This paper presents a new surface content completion framework that can restore both shape and appearance from scanned, incomplete point set inputs. First, the geometric holes can be robustly identified from noisy and defective data sets without the need of any normal or orientation information, using the method of active deformable models. The geometry and texture information of the holes can then be determined either automatically from the models' context, or semi-automatically with minimal users' intervention. The central idea for this repair process is to establish a quantitative similarity measurement among local surface patches based on their local parameterizations and curvature computation. The geometry and texture information of each hole can be completed by warping the candidate region and gluing it to the hole. The displacement for the alignment process is computed by solving a Poisson equation in 2D. Our experiments show that the unified framework, founded upon the techniques of deformable models, local parameterization, and PDE modeling, can provide a robust and elegant solution for content completion of defective, complex point surfaces.

Triangular mesh is one of the most popular shape representations in computer graphics and the CAD/CAM/ CAE area. In this paper we present a memory and time efficient topology construction algorithm from triangle soup, which is a set of triangles without connectivity information. The proposed algorithm consists of the following steps: (1) vertex merging, (2) internal edge linking, (3) multi-disk vertex splitting, and (4) boundary gap stitching. Typical triangle soup comes in the form of an STL file, and topology construction work encounters non-manifold cases for various causes, which should be converted to 2-manifold models for many downstream processes such as rapid prototyping and tool path generation. The proposed algorithm uses a light-weight vertex-based data structure (adapted from a 'corner table' structure), and does not need to construct a full non-manifold topology information. The efficiency of the proposed algorithm is shown by empirical tests on practical examples.

Computation of the characteristic points such as inflection points or cusp on a curve is often necessary in CAGD applications. When a curve is represented in a B-spline form, such computations can be made easier once it is transformed in a set of polynomial curves in a power form. Once a curve is represented in a power form, a point evaluation can be also made faster due to Horner's rule even though some issues of stability remains. In addition, the implicitization process of a parametric curve using a resultant usually requires the geometry represented in a power form. Usual practice of the transformation of a B-spline curve into a set of piecewise polynomial curves in a power form is done by either a knot refinement followed by basis conversions, or applying a Taylor expansion on the B-spline curve for each knot span. Presented in this paper is a new algorithm, called direct expansion, for the problem. The algorithm first locates the coefficients of all the linear terms that make up the basis functions in a knot span, and then the algorithm directly obtains the power form representation of basis functions by expanding the summation of products of appropriate linear terms. Then, a polynomial segment of a knot span can be easily obtained by the summation of products of the basis functions within the knot span with corresponding control points. Repeating this operation for each knot span, all of the polynomials of the B-spline curve can be transformed into a power form.