Real Platform Research Articles

Two-dimensional hourglass Weyl nodal loop in monolayer Pb(ClO2)2 and Sr(ClO2)2

The hourglass fermions in solid-state materials have been attracting significant interest recently. However, realistic two-dimensional (2D) materials with hourglass-shaped band structures are still very scarce. Here, through the first-principles calculations, we identify the monolayer Pb(ClO2)2 and Sr(ClO2)2 materials as the new realistic materials platform to realize 2D hourglass Weyl nodal loop. We show that these monolayer materials possess an hourglass Weyl nodal loop circling around the Γ point and Weyl nodal line on the Brillouin zone (BZ) boundary in the absence of spin–orbit coupling (SOC). Through the symmetry analysis, we demonstrate that the hourglass Weyl nodal loop and Weyl nodal line are protected by the nonsymmorphic symmetries, and are robust under the biaxial strains. When we include the SOC, a tiny gap will be opened in the hourglass nodal loop and nodal line, and the nodal line can be transformed into the spin-orbit Dirac points. Our results provide a new realistic material platform for studying the intriguing physics associated with the 2D hourglass Weyl nodal loop and spin-orbit Dirac points.

BOT-MICS: Bounding Time Using Analytics in Mixed-Criticality Systems

An increasing trend for reducing cost, space, and weight leads to modern embedded systems that execute multiple tasks with different criticality levels on a common hardware platform while guaranteeing a safe operation. In such mixed-criticality (MC) systems, multiple worst case execution times (WCETs) are defined for each task, corresponding to the system operation mode to improve the MC system’s timing behavior at runtime. Determining the appropriate WCETs for lower criticality (LC) modes is nontrivial. On the one hand, considering a very low WCET for tasks can improve the processor utilization by scheduling more tasks in that mode, on the other hand, using a larger WCET ensures that the mode switches (which causes by task overrunning) are minimized, thereby improving the quality of service for all tasks, albeit at the cost of processor utilization. Hitherto, no analytical solutions are proposed to determine WCETs in LC modes. In this regard, we propose a scheme to determine WCETs by the Chebyshev theorem, to make a tradeoff between the number of scheduled tasks at design-time and the number of dropped low-criticality tasks at runtime as a result of frequent mode switches. To have a tight bound of execution times and mode switching probability, we also propose a distribution analytics-based scheme, in which the mode switching probability is obtained based on the cumulative distribution function. Our experimental results show that our scheme improves the utilization of state-of-the-art MC systems by up to 72.27%, while maintaining 24.28% mode switching probability in the worst case scenario. Besides, the results of running embedded real-time benchmarks on a real platform show that the distribution-based scheme can improve the utilization by 7.30% while bounding the mode switching probability by 4.85% more, compared to the Chebyshev-based scheme.

Learning Variable Impedance Control for Aerial Sliding on Uneven Heterogeneous Surfaces by Proprioceptive and Tactile Sensing

The recent development of novel aerial vehicles capable of physically interacting with the environment leads to new applications such as contact-based inspection. These tasks require the robotic system to exchange forces with partially-known environments, which may contain uncertainties including unknown spatially-varying friction properties and discontinuous variations of the surface geometry. Finding a solution that senses, adapts, and remains robust against these environmental uncertainties remains an open challenge. This letter presents a learning-based adaptive control strategy for aerial sliding tasks. In particular, the gains of a standard impedance controller are adjusted in real-time by a neural network policy based on proprioceptive and tactile sensing. This policy is trained in simulation with simplified actuator dynamics in a student-teacher learning setup. The real-world performance of the proposed approach is verified using a tilt-arm omnidirectional flying vehicle. The proposed controller structure combines data-driven and model-based control methods, enabling our approach to successfully transfer directly and without adaptation from simulation to the real platform. We attribute the success of the sim-to-real transfer to the inclusion of feedback control in the training and deployment. We achieved tracking performance and disturbance rejection that cannot be achieved using fine-tuned state of the art interaction control method.

Cash Waqf-Based Crowdfunding Model for Empowering Single Mother Entrepreneurs in Malaysia

This paper aims to explore the potential of cash waqf-based crowdfunding as a source to obtain funds for single mother entrepreneurs. It has been discovered that single mothers are physically, emotionally, and financially struggling to support their lives and perhaps their families. Based on the library research, this study applies the inductive approach and descriptive method in analysing with related works of literature. The finding acknowledges that crowdfunding is frequently argued as a viable alternative to traditional financial institutions. Nevertheless, this approach appears to be different for single mother entrepreneurs, also known as micro-enterprises, because the majority of them lack training, skills and management for attracting investors to support their business. As a result, the findings proposed a unique model of cash waqf-based crowdfunding fund for single mother entrepreneurs. Although this study is based on library research, its findings may not be applicable in terms of the theoretical aspects. Meanwhile, it still creates valuable contributions in accordance of developing a practical model of Islamic crowdfunding particularly cash waqf-based for single mother entrepreneurs in Malaysia. The special model of Islamic crowdfunding that is proposed by this study can potentially support single mother entrepreneurs, if it is applied to a real platform that may reduce the burden of the government bureaucracy because most of them lack financial constraints. Furthermore, this study expands the potential for the application of cash waqf-based crowdfunding as a practical solution for single mother entrepreneurs in Malaysia. Although Islamic crowdfunding is not a new research topic and applications of Islamic crowdfunding as a tool, empirical studies in this area are still limited, specifically investigating the use of Islamic crowdfunding. Despite the fact that this study is conceptual, it provides a practical model which can be utilised as a crowdfunding tool to empower entrepreneurs, especially single mothers.

The impact of formulation of cost function in Task Mapping Problem on NoCs using bio-inspired based-metaheuristics

The Task Mapping Problem (TMP) of a Real-Time Application (RTA) onto a Many-/Multi-Processor System-on-a-Chip (MPSoC) with a Network-on-a-Chip (NoC) as on-chip communication architecture can be tackled as an optimization problem to improve desired design features in a static analysis setting. Examples of such features are the system’s compliance with its time requirements. A search-based bio-inspired metaheuristic such as the Genetic Algorithm (GA) can achieve task placement solutions in which all tasks are schedulable. However, the optimization processes are susceptible to the way the cost function is formulated. Additionally, solutions obtained with some of these methods do not consider some of the platform characteristics, such as the number of available virtual channels in the routers, and memory requirements in the processors, rendering their solutions unsuitable for real platforms. Therefore, we propose four objective functions to unify the end-to-end schedulability analysis that considers such platform characteristics. To demonstrate the performance of our objective functions, we conduct a series of experiments mapping the tasks of a benchmark application together with synthetic generated ones onto multiple Network-on-Chip platforms with realistic characteristics. We compare the objective functions during the optimization process using four different metaheuristics, showing that the multi-objective formulation helps the task mapper to achieve solutions that are more suitable for implemented NoC platforms. Although our study focuses on homogeneous NoC-Based MPSoCs, results have generated a better understanding of how the formulation of cost functions impacts the optimization processes for the specific TMP case, considering the Multi-Objective Optimization (MOO) context.

Attack Detection and Mitigation in MEC-Enabled 5G Networks for AIoT

In recent years, AIoT (Artificial Intelligence of Things) applications have received lots of attention since it exploits widely deployed IoT devices to collect data and perform an action while leveraging AI to obtain knowledge and insights. When deploying AIoT in 5G networks, Multi-access Edge Computing (MEC) is appropriate for enabling local management of IoT devices and computation of machine learning (ML) algorithms. Facing the multitude of threats in the ML data layer, IoT service layer, and 5G communications layer, MEC should enable corresponding detection and mitigation schemes to protect AIoT applications. In this article, we examine the existing security solutions located in each layer and discuss the interrelated challenges. For example, a network-traffic-based IDS solution in MEC might capture IoT malware but fail to identify a growing file-less attack. We suggest that firmware emulation in IoT endpoint should be included to provide system-level behaviors to network-based detectors in MEC so that file-less attacks can be distinguished. The potential of a backdoor attack aiming to poison data or corrupt ML models cannot be ignored in AIoT applications, and a corresponding detector should be implemented in MEC. Due to the rise of low-cost Software-Defined Radio (SDR), malicious attacks using rogue base stations (BSs) have become more popular. It implies that security protection at MEC in the communications layer is necessary. This article, therefore, proposes a novel platform, M3Inspector, where inspectors located in mobiles and AIoT machines collect information from surrounding BSs and provide them to MEC. MEC determines the rogue BS and makes a notification to users subscribing to the local service. A realistic 5G experimental platform with rogue BSs is developed. The results demonstrate that attack detection and mitigation can be implemented in the MEC paradigm to significantly improve the security protection of AIoT from the perspectives of rogue BS attacks and file-less attacks.

Energy efficient mixed task handling on real-time embedded systems using FreeRTOS

With the increased use of embedded devices in Industrial electronics, like relays, battery life has gained more and more attention. Modern processors can use Dynamic Voltage Frequency Scaling (DVFS) techniques for energy reduction and temperature control. However, DVFS is not well supported by systems running with Real-Time operating systems like FreeRTOS, which is a widely used real-time operation system (RTOS) in the industry. Furthermore, energy-efficient DVFS techniques for Mixed Tasksets (including periodic and aperiodic tasks) are hardly investigated. This paper extends the classic well-known DVFS technique Cycle Conserving algorithm to handle Mixed Taskset (CCMT algorithm) and implements it on a real-time embedded platform powered by FreeRTOS. We describe our experience implementing CCMT on a real platform with limited DVFS and corresponding scheduler support. Results show that we can successfully apply CCMT to handle aperiodic requests while meeting the deadlines of the periodic tasks and saving energy on the FreeRTOS platform. The algorithm is tested on an ARM Cortex-M7 processor integrated with frequency scaling and power management. Over 5% to 10% energy savings can be achieved for a standard real-time scheduling mechanism without penalty in application throughput.

Toxicity of trastuzumab for breast cancer spheroids: Application of a novel on-a-chip concentration gradient generator

Evaluating the effectiveness of drugs on an individual basis before prescribing them has an influential role in choosing a successful treatment plan. The advent of microfluidic devices has provided us with more real platforms capable of mimicking the natural microenvironment (scales, shear stress, diffusion transport, etc.). This research has introduced a truncated microfluidic gradient maker that can create a linear concentration gradient of trastuzumab (showing good correlation (r(9) = 0.9968) with simulation data) with a wide range of flow rates (20–190 µl/min) and investigate its effect on SK-BR-3 spheroid cells. The results of trastuzumab efficacy on 3D spheroid cells in the microchip were compared with the conventional 2D cell cultured by MTT assay at 24 h, 48 h, and 72 h. The results showed that trastuzumab had dose-dependent effects in both 2D and spheroid cell cultures. However, the IC50 values for trastuzumab were significantly higher in 2D cell culture (560 ± 23 µg/mL) compared to 3D spheroid models (407 ± 17 µg/mL). Moreover, compared with conventional drug sensitivity tests, this microchip required fewer samples and reagents and a lower risk of fluctuation and contamination due to automatic medium exchange.

Advances and challenges in developing smart, multifunctional microneedles for biomedical applications.

Microneedles (MNs) have been developed as minimally invasive tools for diagnostic and therapeutic applications. However, in recent years, there has been an increasing interest in developing smart multifunctional MN devices to provide automated and closed-loop systems for body fluid extraction, biosensing, and drug delivery in a stimuli-responsive manner. Although this technology is still in its infancy and far from being translated into the clinic, preclinical trials have shown some promise for the broad applications of multifunctional MN devices. The main challenge facing the fabrication of smart MN patches is the integration of multiple modules, such as drug carriers, highly sensitive biosensors, and data analyzers in one miniaturized MN device. Researchers have shown the feasibility of creating smart MNs by integrating stimuli-responsive biomaterials and advanced microscale technologies, such as microsensors and microfluidic systems, to precisely control the transportation of biofluids and drugs throughout the system. These multifunctional MN devices can be envisioned in two distinct strategies. The first type includes individual drug delivery and biosensing MN units with a microfluidic system and a digital analyzer responsible for fluid transportation and communication between these two modules. The second type relies on smart biomaterials that can function as drug deliverers and biosensors by releasing drugs in a stimuli-responsive manner. These smart biomaterials can undergo structural changes when exposed to external stimuli, such as pH and ionic changes, mimicking the biological systems. Studies have demonstrated a high potential of hydrogel-based MN devices for a wide variety of biomedical applications, such as drug and cell delivery, as well as interstitial fluid extraction. Biodegradable hydrogels have also been advantageous for fabricating multifunctional MNs due to their high loading capacity and biocompatibility with the drug of choice. Here, we first review a set of MN devices that can be employed either for biosensing or delivery of multiple target molecules and compare them to the conventional and more simple systems, which are mainly designed for single-molecule sensing or delivery. Subsequently, we expand our insight into advanced MN systems with multiple competencies, such as body fluid extraction, biosensing, and drug delivery at the point of care. The improvement of biomaterials knowledge and biofabrication techniques will allow us to efficiently tune the next generation of smart MNs and provide a realistic platform for more effective personalized therapeutics.

KSN: Modeling and simulation of knowledge using machine learning in NFV/SDN-based networks

The rapid advancement of networking and computing technologies has led to the emergence of a wide range of diverse dynamic network services that present networks with new challenges, including in the provision and management of services. Recently, SDN and NFV technologies can be effective solutions in the dynamic management of network functions. For dealing with dynamic and complex problems in the field of NFV and SDN issues, the use of knowledge-based systems can be a good solution that has received more and more attention from researchers and developers. Despite the increasing attention to the use of knowledge such as machine learning and deep learning methods in the NFV and SDN field, the lack of a simulation framework with the ability to knowledge concept to implement and evaluate new methods has created a gap. In order to evaluate a new method on real platforms, a large-scale infrastructure has to be used, which faces challenges such as availability, cost, time and implementation difficulties. Since the existing simulation frameworks for networks using NFV and SDN technologies (e.g. CloudsimSDNNFV) do not support the knowledge-based concept, to fill this gap, we propose a highly extensible, interoperable, and scalable simulation framework for developing knowledge-based methods such as machine learning algorithms in the field of NFV and SDN, such as virtualized network functions auto-scaling, placement and migration. Also, for the proposed Knowledge-defined SDN/NFV (KSN) framework, an architecture is provided by defining the knowledge plane in the SDN/NFV integration structure. In addition, a new auto-scaling method based on the prediction of NFV resources consumption using machine learning has been proposed and simulated to demonstrate the capabilities of the framework and its evaluation. As we know, this is the first machine learning-based method in the NFV field, including VNF auto-scaling, which has been simulated in a network environment, and the impact of using machine learning has been practically evaluated and observed. The evaluation results show that the proposed method has a better impact on the SLA violation rate and resource usage than ElasticSFC and the basic version (without any auto-scaling method) and therefore it can be concluded that resource prediction is an effective and efficient solution for earlier decision making and VNF auto-scaling.

Split-ring resonator coupling-induced tunable acoustic second-order topological insulators

The recent higher-order topological insulators (HOTIs) encoded with the tight-binding model (TBM) have been demonstrated beyond traditional bulk-edge correspondence, which shows the exotic capability of sound manipulation. However, the current waveguide resonator model with weak tunability in hopping controls has been used as the most common method to construct the TBM directly in a realistic acoustic system. Here, we present another carrier and methodology to realize the TBM based on split-ring resonators (SRRs) coupling in a realistic physical platform and construct a second-order topological insulator (SOTI) based on it, where the hopping and frequencies of the topological corner states can be easily tuned by adjusting the coupling strength of the SRRs. Numerical analysis and experimental measurement demonstrated zero-dimensional topological corner states in this acoustic resonant system. The findings provide a more flexible method for tunable acoustic HOTIs realization.

Application of Neural Data Processing in Autonomous Model Platform—A Complex Review of Solutions, Design and Implementation

One of the bottlenecks of autonomous systems is to identify and/or design models and tools that are not too resource demanding. This paper presents the concept and design process of a moving platform structure–electric vehicle. The objective is to use artificial intelligence methods to control the model’s operation in a resource scarce computation environment. Neural approaches are used for data analysis, path planning, speed control and implementation of the vision system for road sign recognition. For this purpose, multilayer perceptron neural networks and deep learning models are used. In addition to the neural algorithms and several applications, the hardware implementation is described. Simulation results of systems are gathered, data gathered from real platform tests are analyzed. Experimental results show that low-cost hardware may be used to develop an effective working platform capable of autonomous operation in defined conditions.

Sim-to-Real: Mapless Navigation for USVs Using Deep Reinforcement Learning

In recent years, mapless navigation using deep reinforcement learning algorithms has shown significant advantages in improving robot motion planning capabilities. However, the majority of past works have focused on aerial and ground robotics, with very little attention being paid to unmanned surface vehicle (USV) navigation and ultimate deployment on real platforms. In response, this paper proposes a mapless navigation method based on deep reinforcement learning for USVs. Specifically, we carefully design the observation space, action space, reward function, and neural network for a navigation policy that allows the USV to reach the destination collision-free when equipped with only local sensors. Aiming at the sim-to-real transfer and slow convergence of deep reinforcement learning, this paper proposes a dynamics-free training and consistency strategy and designs domain randomization and adaptive curriculum learning. The method was evaluated using a range of tests applied to simulated and physical environments and was proven to work effectively in a real navigation environment.

Efficient Dialog Policy Learning by Reasoning with Contextual Knowledge

Goal-oriented dialog policy learning algorithms aim to learn a dialog policy for selecting language actions based on the current dialog state. Deep reinforcement learning methods have been used for dialog policy learning. This work is motivated by the observation that, although dialog is a domain with rich contextual knowledge, reinforcement learning methods are ill-equipped to incorporate such knowledge into the dialog policy learning process. In this paper, we develop a deep reinforcement learning framework for goal-oriented dialog policy learning that learns user preferences from user goal data, while leveraging commonsense knowledge from people. The developed framework has been evaluated using a realistic dialog simulation platform. Compared with baselines from the literature and the ablations of our approach, we see significant improvements in learning efficiency and the quality of the computed action policies.

LIMITLESS — LIght-weight MonItoring Tool for LargE Scale Systems

This work presents LIMITLESS, a HPC framework that provides new strategies for monitoring clusters. LIMITLESS is a scalable light-weight monitor that is integrated with other HPC runtimes in order to obtain a holistic view of the system that combines both platform and application monitoring. This paper presents a description of the novel components of the architecture, including new approaches for reaching a higher scalability based on a combination of in-transit processing and performance prediction. We also include a methodology for improving application scheduling by means of machine learning classifiers and application profiling. This work also includes a practical evaluation on simulated and real platforms, that shows significant monitoring scalability, retrieving data capacity and reduced overheads. Results show that the performance prediction techniques reduce communications and the number of monitoring packets by more than 90% on average, and the fine-grain scheduling allows LIMITLESS to run applications in shared nodes reducing the makespan by 25% and saving resources.

Abstract 4088: A novel real world clinico-genomics data platform identifies breast cancer histologic and molecular characteristics as candidate predictors for time to distant metastasis

Abstract Introduction: Currently, the NCCN guidelines for invasive breast cancer (BC) surveillance are the same regardless of tumor pathology. The length of time between initial diagnosis and documented distant metastasis based on tumor subtypes has not been well-studied in real-world settings, which means that for any given patient, generalized screening recommendations may be either too stringent or too relaxed. Here, we report how tumor histologic and molecular characteristics analyzed from a real-world data platform correlate with the average time to distant recurrence. This information can guide more personalized surveillance for BC patients. Methods: Ciitizen (a wholly owned subsidiary of Invitae) leverages HIPAA right of access on behalf of patients to collect their medical records, extracts relevant clinical information into a standardized and validated longitudinal dataset, and - with their consent - enables research use of this data. From our current invasive BC cohort (n=1011), we identified a subcohort who were diagnosed at Stage I-III with associated histology and/or molecular characterization, then ultimately went on to develop distant metastases (n=195 cases involving 321 sites: bone, brain, liver, and/or lung). We calculated time from initial diagnosis to confirmation of distant metastasis at each site, and statistically analyzed differences using pooled hypothesis testing. Results: In addition to finding expected associations between molecular type and site(s) of metastasis, i.e. bone and brain more common in Hormone Receptor Positive (HR+) subtypes vs HR-, we also characterized the average time to distant metastasis (TDM) to identify other statistically significant associations: Molecular Type: HR-/HER2- (n=30) has an average TDM of 1124 days, which is significantly faster than HR+/HER2- (2224 days; n=197) and HR+/HER2+ (2159 days; n=32) (p=0.0043). BRCA: Individuals with BRCA1/2 pathogenic germline variants have an average TDM of 1043 days (n=15), significantly earlier than individuals with normal BRCA1/2 (2238 days; n=179) (p <0.0001). Histology: Ductal carcinomas have an average TDM of 2262 days (n=242) and lobular carcinomas have an average TDM of 2812 days (n=24), both of which are significantly longer than the rarer metaplastic (739 days; n=11) and inflammatory (978 days; n=5) (p=0.0001). Conclusions: These real-world findings characterize average time to metastasis and reveal statistically significant associations based on histologic and molecular type, which taken together can inform more personalized screening recommendations for BC patients. Since the Ciitizen data platform also includes additional treatment and outcome variables, future studies can explore the relationships between these various features towards more granular prognostic and predictive classifications. Citation Format: Amanda Nottke, Anthony B. Cardillo, Hana E. Littleford, Lisandra West-Odell, Alexandra Berk. A novel real world clinico-genomics data platform identifies breast cancer histologic and molecular characteristics as candidate predictors for time to distant metastasis [abstract]. In: Proceedings of the American Association for Cancer Research Annual Meeting 2022; 2022 Apr 8-13. Philadelphia (PA): AACR; Cancer Res 2022;82(12_Suppl):Abstract nr 4088.

State estimation with trusted phasor measuring unit placements of distribution system involving renewable energy sources

Among Renewable Energy Sources wind and solar energy are the most prominent and favorable substitutes to meet mankind’s future electricity requirements. Typically, these resources are combined in a distribution network to provide local distribution users. The amalgamation of these upgrades into the distribution network may modify the malfunction and network topologies, which may fail due to their pre-set state. Therefore, an advanced and accurate tracking device must continuously monitor topology changes, which is the Phasor Measuring Unit (PMU). This work proposes a new technique based on the Sea Lion Optimization Algorithm to determine the Optimal PMU Placements and its employment positions, i.e. the power structure is completely perceptible. In addition, the cost of system losses can be calculated with or without wind energy to achieve energy reserves. The energy losses before and after the wind turbine connection is compared to realize energy reserves. This comparison of energy reserves is made using Realistic and Stochastic platforms. Comparison and proposed formulas offered in the IEEE 15, 33, 69, and 85 bus distribution networks were explored to prove their effectiveness. Moreover, the proposed methodology produces more trusted results than those of other methods in the literature.

Application of Digital Technology in Intangible Cultural Heritage Protection

Intangible cultural heritage itself is fragile, so it needs human protection. The existence of everything is a dynamic process, and its emergence, growth, continuation, and demise are objective existences, including “intangible cultural heritage.” With the continuous deepening and development of science and technology, the development of digital acquisition technology, storage technology, remote sensing technology, and spatial modeling technology provides a corresponding realistic basis and platform for the protection of different types of intangible cultural heritage. This paper mainly discusses the application of a variety of modern information technologies in the protection of intangible cultural heritage, deeply discusses the problems existing in digital protection of intangible cultural heritage, analyzes the characteristics of digital protection of intangible cultural heritage, discusses how to better establish the protection platform and management mechanism, and puts forward the application prospect of computer digital technology in the protection of intangible cultural heritage, so as to realize relatively comprehensive digital protection of intangible cultural heritage.

CARTAD: Compiler-Assisted Reinforcement Learning for Thermal-Aware Task Scheduling and DVFS on Multicores

As the power density of modern CPUs is gradually increasing, thermal management has become one of the primary concerns for multicore systems, where task scheduling and dynamic voltage/frequency scaling (DVFS) play a pivotal role in effectively managing the system temperature. In this article, we propose <i>CARTAD</i>, a new reinforcement learning (RL)-based task scheduling and DVFS method for temperature minimization and latency guarantee on multicore systems. The novelty of <i>CARTAD</i> framework is that we exploit the machine learning technique to analyze the applications&#x2019; intermediate representations (IRs) generated by a compiler and identify an important feature which is critical for predicting the application&#x2019;s performance. With the newly explored feature, we construct an RL-based scheduler with the more effective state representation and reward function such that the system temperature can be minimized while guaranteeing applications&#x2019; latency. We implement and evaluate <i>CARTAD</i> on real platforms in comparison with the state-of-the-art approaches. Experimental results show <i>CARTAD</i> can reduce the maximum temperature by up to 16 &#x00B0;C and the average temperature by up to 10 &#x00B0;C.

Meal delivery routing optimization with order allocation strategy based on transfer stations for instant logistics services

The rapid development of instant logistics services has brought not only convenience to people's life but also a great challenge to traffic management. Due to the limited delivery range of instant delivery systems, customers are usually recommended the meals nearby or pay much higher delivery fees for long-distance delivery. This study proposes a novel order splitting strategy based on transfer stations for long-distance meal orders to satisfy the diverse customer demands. The meal delivery routing problem is addressed with an order allocation strategy based on transfer stations through developing a three-stage modelling framework consisting of order combination, splitting and delivery routing for the online-to-offline instant logistics services. Normal meal orders are combined by the DBSCAN algorithm, and the cross-regional long-distance orders are split by transfer stations. Based on order combination and splitting, a mixed integer programming model is constructed for the meal delivery routing problem and solved by the adaptive large neighbourhood search algorithm. The proposed algorithm converges quickly for the tested instances constructed based on real platform data. The proposed order allocation strategy can expand the delivery scopes of couriers effectively, stimulate more potential orders and guarantee the timeliness of meal delivery.