Transport Algorithm Research Articles

Unleashing the power of artificial intelligence in phonon thermal transport: Current challenges and prospects

The discovery of advanced thermal materials with exceptional phonon properties drives technological advancements, impacting innovations from electronics to superconductors. Understanding the intricate relationship between composition, structure, and phonon thermal transport properties is crucial for speeding up such discovery. Exploring innovative materials involves navigating vast design spaces and considering chemical and structural factors on multiple scales and modalities. Artificial intelligence (AI) is transforming science and engineering and poised to transform discovery and innovation. This era offers a unique opportunity to establish a new paradigm for the discovery of advanced materials by leveraging databases, simulations, and accumulated knowledge, venturing into experimental frontiers, and incorporating cutting-edge AI technologies. In this perspective, first, the general approach of density functional theory (DFT) coupled with phonon Boltzmann transport equation (BTE) for predicting comprehensive phonon properties will be reviewed. Then, to circumvent the extremely computationally demanding DFT + BTE approach, some early studies and progress of deploying AI/machine learning (ML) models to phonon thermal transport in the context of structure–phonon property relationship prediction will be presented, and their limitations will also be discussed. Finally, a summary of current challenges and an outlook of future trends will be given. Further development of incorporating AI/ML algorithms for phonon thermal transport could range from phonon database construction to universal machine learning potential training, to inverse design of materials with target phonon properties and to extend ML models beyond traditional phonons.

Improved object detection method for unmanned driving based on Transformers

Improved object detection method for unmanned driving based on Transformers

SRFDet3D: Sparse region fusion based 3D object detection

Unlike the earlier 3D object detection approaches that formulate hand-crafted dense (in thousands) object proposals by leveraging anchors on dense feature maps, we formulate np (in hundreds) number of learnable sparse object proposals to predict 3D bounding box parameters. The sparse proposals in our approach are not only learnt during training but also are input-dependent, so they represent better object candidates during inference. Leveraging the sparse proposals, we fuse only the sparse regions of multi-modal features and we propose Sparse Region Fusion based 3D object Detection (SRFDet3D) network with mainly three components: an encoder for feature extraction, a region proposal generation module for sparse input-dependent proposals and a decoder for multi-modal feature fusion and iterative refinement of object proposals. Additionally for optimal training, we formulate our sparse detector with many-to-one label assignment based on Optimal Transport Algorithm (OTA). We conduct extensive experiments and analysis on publicly available large-scale autonomous driving datasets: nuScenes, KITTI, and Waymo. Our LiDAR-only SRFDet3D-L network achieves 63.1 mAP and outperforms the state-of-the-art networks on the nuScenes dataset, surpassing the dense detectors on KITTI and Waymo datasets. Our LiDAR-Camera model SRFDet3D achieves 64.7 mAP with improvements over existing fusion methods.

Preserving Tracer Correlations in Moment‐Based Atmospheric Transport Models

AbstractA linear non‐diffusive algorithm for advective transport is developed that greatly improves the detail at which aerosols and clouds can be represented in atmospheric models. Linear advection schemes preserve tracer correlations but the most basic linear scheme is rarely used by atmospheric modelers on account of its excessive numerical diffusion. Higher‐order schemes are in widespread use, but these present new problems as nonlinear adjustments are required to avoid occurrences of negative concentrations, spurious oscillations, and other non‐physical effects. Generally successful at reducing numerical diffusion during the advection of individual tracers, for example, particle number or mass, the higher‐order schemes fail to preserve even the simplest of correlations between interrelated tracers. As a result, important attributes of aerosol and cloud populations including radial moments of particle size distributions, molecular precursors related through chemical equilibria, aerosol mixing state, and distribution of cloud phase are poorly represented. We introduce a new transport scheme, minVAR, that is both non‐diffusive and preservative of tracer correlations, thereby combining the best features of the basic and higher‐order schemes while enabling new features such as the tracking of sub‐grid information at arbitrarily fine scales with high computational efficiency.

A dynamic compartment model for spatially heterogeneous reactors: Scalar and Monte-Carlo particle mixing

The coupling of multidimensional population balance models with transport equations is often necessary for the simulation of spatially heterogeneous reactors. Because of numerical intractability, a compromise on both the hydrodynamics and the population description has to be made. In this work, the Monte-Carlo method is used to calculate the local integral source term due to the dispersed phase (isodense particles) and a compartment model is used to describe the fluid motion in the reactor. The number density function n(x,t,ξ) is approximated by the number of Monte-Carlo particles per compartment and their movement in the space of compartiment uses a stochastic transport algorithm adapted from existing works. The numerical tool is thoroughly and successfully validated in the context of mixing and first order reaction in a 70L stirred tank equipped with a Rushton turbine. The dynamic method makes use of time resolved CFD simulations to built time dependent flow matrices reflecting the periodic nature of the flow field. The dynamic approach is finally compared to other usual compartment model approaches in terms of complexity, accuracy and rapidity which reveals the benefit of considering the flow unsteadiness in a compartimented stirred tank.

Optimal Transport with Cyclic Symmetry

We propose novel fast algorithms for optimal transport (OT) utilizing a cyclic symmetry structure of input data. Such OT with cyclic symmetry appears universally in various real-world examples: image processing, urban planning, and graph processing. Our main idea is to reduce OT to a small optimization problem that has significantly fewer variables by utilizing cyclic symmetry and various optimization techniques. On the basis of this reduction, our algorithms solve the small optimization problem instead of the original OT. As a result, our algorithms obtain the optimal solution and the objective function value of the original OT faster than solving the original OT directly. In this paper, our focus is on two crucial OT formulations: the linear programming OT (LOT) and the strongly convex-regularized OT, which includes the well-known entropy-regularized OT (EROT). Experiments show the effectiveness of our algorithms for LOT and EROT in synthetic/real-world data that has a strict/approximate cyclic symmetry structure. Through theoretical and experimental results, this paper successfully introduces the concept of symmetry into the OT research field for the first time.

Optimization-based, property-preserving algorithm for passive tracer transport

We present a new optimization-based property-preserving algorithm for passive tracer transport. The algorithm utilizes a semi-Lagrangian approach based on incremental remapping of the mass and the total tracer. However, unlike traditional semi-Lagrangian schemes, which remap the density and the tracer mixing ratio through monotone reconstruction or flux correction, we utilize an optimization-based remapping that enforces conservation and local bounds as optimization constraints. In so doing we separate accuracy considerations from preservation of physical properties to obtain a conservative, second-order accurate transport scheme that also has a notion of optimality. Moreover, we prove that the optimization-based algorithm preserves linear relationships between tracer mixing ratios. We illustrate the properties of the new algorithm using a series of standard tracer transport test problems in a plane and on a sphere.

Investigation of ionization chamber perturbation factors using proton beam and Fano cavity test for the Monte Carlo simulation code PHITS.

The reference dose for clinical proton beam therapy is based on ionization chamber dosimetry. However, data on uncertainties in proton dosimetry are lacking, and multifaceted studies are required. Monte Carlo simulations are useful tools for calculating ionization chamber dosimetry in radiation fields and are sensitive to the transport algorithm parameters when particles are transported in a heterogeneous region. We aimed to evaluate the proton transport algorithm of the Particle and Heavy Ion Transport Code System (PHITS) using the Fano test. The response of the ionization chamber and beam quality correction factors were calculated using the same parameters as those in the Fano test and compared with those of other Monte Carlo codes for verification. The geometry of the Fano test consisted of a cylindrical gas-filled cavity sandwiched between two cylindrical walls. was calculated as the ratio of the absorbed dose in water to the dose in the cavity in the chamber. We compared the calculated using PHITS with that of a previous study, which was calculated using other Monte Carlo codes (Geant4, FULKA, and PENH) under similar conditions. The flight mesh, a parameter for charged particle transport, passed the Fano test within 0.15%. This was shown to be sufficiently accurate compared with that observed in previous studies. The calculated using PHITS were 1.116 ± 0.002 and 1.124 ± 0.003 for NACP-02 and PTW-30013, respectively, and the were 0.981 ± 0.008 and 1.027 ± 0.008, respectively, at 150MeV. Our results indicate that PHITS can calculate the and with high precision.

On the geometry and dynamical formulation of the Sinkhorn algorithm for optimal transport

On the geometry and dynamical formulation of the Sinkhorn algorithm for optimal transport

Globally Optimal Assignment Algorithm for Collective Object Transport Using Air–Ground Multirobot Teams

Globally Optimal Assignment Algorithm for Collective Object Transport Using Air–Ground Multirobot Teams

SUSTAINABLE FOOD PACKAGING IMPACT TO THE REDUCTION OF TRANSPORT COSTS

People dealt with many new issues related to sustainable development goals. Different decisions oriented to sustainable food packing would be very useful, if they could bring significant effect through the reduction of CO2 emission. The paper focuses on sustainable food packing and its effect on reduction of freight transportation costs. This topic has many research gaps and requires theoretical and practical investigations. Integration of sustainability into long term process has clear implications for the environment. Meherishi et al. (2019) highlight that the process of creating sustainable packaging guarantees the optimal use of materials and energy and is efficiently regenerated without wasting of natural resources. The impact of food packaging on transportation is clear and depends on a number of factors, which are shaped by the characteristics of the food packaging. In addition, among benefits of sustainable food packaging are such main functions as product safety and its identification. Modern complex solutions covering sustainable packing and transport costs reduction are important for increasing environmental sustainability. The study aims to analyze the impact of packaging on the reduction of transport costs. The aim was reached by performing sustainable food packaging and its impact on the reduction of transport costs theoretical justification and by conducting biometric analysis. The overview of the literature of sustainable food packaging and its impact on the supply chain was conducted, and it was found that the food packaging described different types of packaging and new packaging technologies which have an impact on the supply chain and are determined by some factors related to the characteristics of food packaging and the nature of the supply chain. Equally the overview of literature on the transport of food dynamics and the role of transport costs reduction on sustainability was also conducted. It was found that there is a need to develop new technologies and methods for food management and CO2 emissions for longer transport routes reduction. Either the research was conducted by biometric analysis for 2022-2023 to identify current trends, and it was found that the number of analyzed articles, included four clusters, which show that topics are oriented to delivery, flow, algorithms, structure, and role. The bibliometry analysis results show that product packing has links with transport and delivery and algorithms. The authors provided investigations under the topic as it requires knowledge to fill the existing gaps.

AAformer: Auto-Aligned Transformer for Person Re-Identification.

In person re-identification (re-ID), extracting part-level features from person images has been verified to be crucial to offer fine-grained information. Most of the existing CNN-based methods only locate the human parts coarsely, or rely on pretrained human parsing models and fail in locating the identifiable nonhuman parts (e.g., knapsack). In this article, we introduce an alignment scheme in transformer architecture for the first time and propose the auto-aligned transformer (AAformer) to automatically locate both the human parts and nonhuman ones at patch level. We introduce the "Part tokens (PARTs)", which are learnable vectors, to extract part features in the transformer. A PART only interacts with a local subset of patches in self-attention and learns to be the part representation. To adaptively group the image patches into different subsets, we design the auto-alignment. Auto-alignment employs a fast variant of optimal transport (OT) algorithm to online cluster the patch embeddings into several groups with the PARTs as their prototypes. AAformer integrates the part alignment into the self-attention and the output PARTs can be directly used as part features for retrieval. Extensive experiments validate the effectiveness of PARTs and the superiority of AAformer over various state-of-the-art methods.

Towards Quantum Ray Tracing.

Rendering on conventional computers is capable of generating realistic imagery, but the computational complexity of these light transport algorithms is a limiting factor of image synthesis. Quantum computers have the potential to significantly improve rendering performance through reducing the underlying complexity of the algorithms behind light transport. This paper investigates hybrid quantum-classical algorithms for ray tracing, a core component of most rendering techniques. Through a practical implementation of quantum ray tracing in a 3D environment, we show quantum approaches provide a quadratic improvement in query complexity compared to the equivalent classical approach. Based on domain specific knowledge, we then propose algorithms to significantly reduce the computation required for quantum ray tracing through exploiting image space coherence and a principled termination criteria for quantum searching. We show results obtained using a simulator for both Whitted style ray tracing, and for accelerating ray tracing operations when performing classical Monte Carlo integration for area lights and indirect illumination.

Efficient algorithms for Monte Carlo particle transport on AI accelerator hardware

The recent trend toward deep learning has led to the development of a variety of highly innovative AI accelerator architectures. One such architecture, the Cerebras Wafer-Scale Engine 2 (WSE-2), features 40 GB of on-chip SRAM, making it a potentially attractive platform for latency- or bandwidth-bound HPC simulation workloads. In this study, we examine the feasibility of performing continuous energy Monte Carlo (MC) particle transport on the WSE-2 by porting a key kernel from the MC transport algorithm to Cerebras's CSL programming model. New algorithms for minimizing communication costs and for handling load balancing are developed and tested. The WSE-2 is found to run 130 times faster than a highly optimized CUDA version of the kernel run on an NVIDIA A100 GPU—significantly outpacing the expected performance increase given the difference in transistor counts between the architectures.

Explicit and implicit large eddy simulations of variable-density low-speed flows

The application of the Non-oscillatory Forward-in-Time (NFT) class of solvers to facilitate conventional explicit Large Eddy Simulations (LES) and implicit LES (ILES) of low-speed turbulent flow encompassing substantial local density variations is explored. Schemes solving two different sets of Low Mach Number (LMN) equations are proposed with the aim of capturing variable-density effects arising from either a non-isothermal distribution in single-species flow or compositional variation occurring in a binary-species mixing under isothermal conditions. Both schemes employ the semi-implicit NFT integrators based on the Multidimensional Positive Definite Advection Transport Algorithm and a non-symmetric Krylov subspace solver for the variable-coefficient Poisson problems arising from different treatments applied to the non-zero velocity divergence term in the two sets of the LMN equations. The developments are successfully validated using three diverse test cases, differentially heated cavity flow, non-isothermal free-jets, and a helium plume. The reported simulations show that both the dynamic Smagorinsky model in LES and ILES subgrid-scale treatments yield similar accuracy in capturing turbulent effects for the considered flows.

Reconstructing growth and dynamic trajectories from single-cell transcriptomics data

Time-series single-cell RNA sequencing (scRNA-seq) datasets provide unprecedented opportunities to learn dynamic processes of cellular systems. Due to the destructive nature of sequencing, it remains challenging to link the scRNA-seq snapshots sampled at different time points. Here we present TIGON, a dynamic, unbalanced optimal transport algorithm that reconstructs dynamic trajectories and population growth simultaneously as well as the underlying gene regulatory network from multiple snapshots. To tackle the high-dimensional optimal transport problem, we introduce a deep learning method using a dimensionless formulation based on the Wasserstein–Fisher–Rao (WFR) distance. TIGON is evaluated on simulated data and compared with existing methods for its robustness and accuracy in predicting cell state transition and cell population growth. Using three scRNA-seq datasets, we show the importance of growth in the temporal inference, TIGON’s capability in reconstructing gene expression at unmeasured time points and its applications to temporal gene regulatory networks and cell–cell communication inference.

A dose calculation algorithm for boron neutron capture therapy using convolution/superposition method

The convolution/superposition (C/S) method originally designed for photon dose calculation was first applied for developing a treatment planning system for boron neutron capture therapy. The original concept of TEGMA (total energy generated per unit mass) was proposed to represent distinctive dose components from neutron reactions with the elements in the patient's tissue. First, neutron fluence distributions in a homogeneous brain phantom irradiated with an energy-groupwise pencil beam of 2.5 × 2.5 mm2 were calculated using the MCNP6.2 code. Then, a library of energy-groupwise TEGMA and KERMA were generated and stored in the developed C/S code. As a benchmark, dose distributions in a cuboid phantom and a human head phantom were calculated using the developed C/S and PHITS Monte Carlo codes. A neutron beam having a continuous epithermal spectrum and a square field of 22.5 × 22.5 mm2 or a circle field of 22.5 mm in diameter was assumed to be incident on the phantoms. The human head phantom was created by the pre-processing including the voxelization and transformation of test DICOM CT images. The differences in boron doses between C/S and MC ranged from 2% to 6%. In nitrogen doses, the differences were from 4% to 9%. A large discrepancy observed in hydrogen lateral dose profiles could be explained by the differences in cross-section data and recoil-proton transport algorithms of MCNP6.2 and PHITS. With isodose curves normalized at the center of the tumor in the human head phantom, they were almost identical in the range of 60%–110% for both cases. The C/S have underestimated the backscattering neutron and showed a larger absorbed dose gradient around 40% region. The calculation time of C/S using Intel i7-10700 processor was less than 1 min for both phantoms. The calculation time of PHITS using three Intel Xeon E5-2640 v4 processors was 15.5 min for the cuboid phantom and ∼380 min for the human head phantom. The proposed algorithm has the advantages of high speed while promising fair accuracy in BNCT dose calculations.

Stochastic simulation of electron transport in a strong electrical field in low-dimensional heterostructures

Abstract In this paper we develop a stochastic simulation algorithm for electron transport in a DA-pHEMT heterostructure. Mathematical formulation of the problem of electron gas transport in the heterostructure in the form of a coupled system of Poisson, Schrödinger and kinetic Boltzmann equations is given. A Monte Carlo model of electron transport in DA-pHEMT heterostructures which accounts for multivalley parabolic band structure, as well as relevant formulas for calculating electron scattering rates and scattering phase functions on polar optical, intervalley phonons and on impurities are developed. The results of a computational experiment involving the solution of the system of Poisson–Schrödinger–Boltzmann equations for the AlGaAs/GaAs/InGaAs/GaAs/AlGaAs heterostructure are presented. The distribution of electrons by energy subband in the main and satellite valleys and the field dependences of the electron drift velocity in each valley are calculated. It was discovered that there is no spatial transfer of electrons into wide-gap AlGaAs layers due to high barriers created by modulated-doped impurities. A comparative analysis of the electron drift velocities in the studied DA-pHEMT heterostructures and in the unstrained layer of the InGaAs is given.

The GenCol Algorithm for High-Dimensional Optimal Transport: General Formulation and Application to Barycenters and Wasserstein Splines

The GenCol Algorithm for High-Dimensional Optimal Transport: General Formulation and Application to Barycenters and Wasserstein Splines

Coarse-Grained Model for Prediction of Hole Mobility in Polyethylene.

Electrical conductivity measurements of polyethylene indicate that the semicrystalline structure and morphology influence the conductivity. To include this effect in atomistic charge transport simulations, models that explicitly or implicitly take morphology into account are required. In the literature, charge transport simulations of amorphous polyethylene have been successfully performed using short oligomers to represent the polymer. However, a more realistic representation of the polymer structure is desired, requiring the development of fast and efficient charge transport algorithms that can handle large molecular systems through coarse-graining. Here, such a model for charge transport simulations in polyethylene is presented. Quantum chemistry calculations were used to define six segmentation rules on how to divide a polymer chain into shorter segments representing localized molecular orbitals. Applying the rules to amorphous systems yields distributions of segments with mode and median segment lengths relatively close to the persistence length of polyethylene. In an initial test, the segments of an amorphous polyethylene were used as hopping sites in kinetic Monte Carlo (KMC) simulations, which yielded simulated hole mobilities that were within the experimental range. The activation energy of the simulated system was lower compared to the experimental values reported in the literature. A conclusion may be that the experimental result can only be explained by a model containing chemical defects that generate deep traps.