Context-aware Services Research Articles

Enhancing intelligent transport systems: A cutting-edge framework for context-aware service management with hybrid deep learning

This study presents a comprehensive framework for optimizing intelligent transport systems (ITS) by integrating advanced communication and information technologies into vehicles, roads, and infrastructure. The primary goal is to enhance transportation efficiency, safety, and environmental sustainability while improving overall mobility for people and goods. Leveraging contextual information, the framework offers personalized, proactive services such as real-time traffic updates, route recommendations, and parking availability. Additionally, it enhances safety and security by providing early hazard warnings and adapting to changing road conditions. Our proposed framework utilizes the enhanced coral reef optimization (ECRO) algorithm to efficiently group vehicles for energy-saving data collection, maximizing information gathering efficiency. Collected data is then transmitted to a central data gathering center via a sink node optimized through the modified pelican optimization (MPO) algorithm, considering various vehicle node design constraints. An incident detection module accurately classifies and detects road incidents, enabling timely emergency service requests and alternate route recommendations. To facilitate incident detection, we introduce the deep Rigdelet neural network (DRNN), a novel deep learning technique tailored for decision-making in incident classification. We validate our framework's performance through NS-2 simulations using the SUMO traffic generator, demonstrating its effectiveness in meeting quality of service (QoS) metrics. Through comparative analysis with existing frameworks, our proposed approach stands out for its superior performance and ability to optimize ITS operations.

Design of a shared context-aware assistive robot for personalised aged-care services

ABSTRACT Background: Sensors, robotics, the Internet of Things (IoT) and data analytics hold considerable potential for enhancing the efficiency of care and health outcomes in the health and aged care sectors. The Australian aged care sector in particular faces significant challenges due to an aging population and outdated technological infrastructure. This article builds upon the foundational research conducted through the collaboration between the University of New South Wales and the University of the Sunshine Coast, as part of the European Union VictoryaHome (VH) project (2014–2016). The VH project, which involved partners from Australia, Norway, Sweden, the Netherlands, and Portugal, aimed to enhance the emotional well-being and security of elderly individuals in their homes. The present study specifically addresses some of the challenges identified in relation to the prohibitively high costs and usability of robots for personalized services in the aged care sector. Methods: This paper presents the design and implementation of an affordable service robot specifically developed to offer personalized services to multiple users in aged care facilities. The system utilizes a Raspberry Pi 4 and incorporates functionalities that are tailored to the needs identified through interviews with facility managers and staff. Results: We successfully demonstrate tracking of entities within a room-scale environment and the switching of profiles using cost-effective devices. Moreover, by tracking multiple entities simultaneously we were able to achieve powerful context-aware functionality that was seamless to the end-user. Conclusion: This study underscores the feasibility and potential benefits of employing low-cost, context-aware service robots in aged care environments.

Smart Resource Allocation in Mobile Cloud Next-Generation Network (NGN) Orchestration with Context-Aware Data and Machine Learning for the Cost Optimization of Microservice Applications

Mobile cloud computing (MCC) provides resources to users to handle smart mobile applications. In MCC, task scheduling is the solution for mobile users’ context-aware computation resource-rich applications. Most existing approaches have achieved a moderate service reliability rate due to a lack of instance-centric resource estimations and task offloading, a statistical NP-hard problem. The current intelligent scheduling process cannot address NP-hard problems due to traditional task offloading approaches. To address this problem, the authors design an efficient context-aware service offloading approach based on instance-centric measurements. The revised machine learning model/algorithm employs task adaptation to make decisions regarding task offloading. The proposed MCVS scheduling algorithm predicts the usage rates of individual microservices for a practical task scheduling scheme, considering mobile device time, cost, network, location, and central processing unit (CPU) power to train data. One notable feature of the microservice software architecture is its capacity to facilitate the scalability, flexibility, and independent deployment of individual components. A series of simulation results show the efficiency of the proposed technique based on offloading, CPU usage, and execution time metrics. The experimental results efficiently show the learning rate in training and testing in comparison with existing approaches, showing efficient training and task offloading phases. The proposed system has lower costs and uses less energy to offload microservices in MCC. Graphical results are presented to define the effectiveness of the proposed model. For a service arrival rate of 80%, the proposed model achieves an average 4.5% service offloading rate and 0.18% CPU usage rate compared with state-of-the-art approaches. The proposed method demonstrates efficiency in terms of cost and energy savings for microservice offloading in mobile cloud computing (MCC).

Enhancing Flexibility in Industry 4.0 Workflows: A Context-Aware Component for Dynamic Service Orchestration

Manufacturing processes of the future will rely on standards for asset interoperability and service orchestration. The Asset Administration Shell (AAS) facilitates information exchange among Industry 4.0 assets, while standardized Business Processes enable workflow execution in manufacturing systems. Combining these technologies provides agility and scalability to manufacturing systems by incorporating asset services within business processes. Service orchestration involves coordinating multiple services, which must be dynamic during runtime to manage unforeseen situations that may arise during the manufacturing process. Context information plays a crucial role in identifying such scenarios and selecting the most suitable devices/services in response, and the Semantic Web accurately represents this information. This paper proposes a context-aware approach for service orchestration using industrial asset services. Our contributions include (1) a component for Context-Aware Service Re-Selection. (2) a domain-specific ontology (DeviceServiceOnt) for Semantic Web-based context representation. And, (3) validation of our proposal in a manufacturing setting where robots are responsible for dispatching and distributing materials within a warehouse. Opportunities for future work are also highlighted, with a primary focus on enhancing workflow dynamicity with context-aware capabilities.

A survey of indoor positioning systems based on a six-layer model

Indoor positioning has attracted considerable interest in both the industry and academic communities because of its wide range of applications, such as asset tracking, healthcare and context-aware services like targeted advertisements. While there are many indoor localisation methods, each has its advantages and disadvantages, taking into consideration various factors such as the effect of the indoor environment, ease of implementation, computational cost, positioning accuracy, etc. In other words, no single solution can cater for all different situations. Although many survey papers have been published on indoor positioning, new techniques and methods are proposed every year, so it is important to stay abreast of its latest developments. In addition, each survey has its own classification for indoor positioning systems without a common scheme. Inspired by the well-known OSI model and TCP/IP model, it would be desirable to develop a systematic framework for studying indoor positioning systems. In this paper, we make this new contribution by introducing a systemic survey framework based on a six-layer model to give a comprehensive survey of indoor positioning systems, namely: device layer, communication layer, network layer, data layer, method layer and application layer. Complementing the previous survey papers, this paper provides a survey of the latest research works on indoor positioning based on the six-layer model. Our emphasis is on systematic categorisation, machine learning-based enhancements, collaborative localisation and COVID-19-related applications. The six-layer model should provide a useful framework and new insights for the research community.

Power consumption and energy management for edge computing: state of the art

Edge computing aims to make internet-based services and remote computing power close to the user by placing information technology (IT) infrastructure at the network edges. This proximity provides data centers with low-latency and context-aware services. Edge computing power consumption is mainly caused by data centers, network equipment, and user equipment. With edge computing (EC), energy management platforms for residential, industrial, and commercial sectors are built. Energy efficiency is considered to be one of the key aspects of edge power constraints. This paper provides the state of the art of power consumption and energy management for edge computing, the computation offloading methods, and more important highlights the power efficiency of edge computing systems. Furthermore, renewable energy and related concepts will also be explored and presented since no human participation is required in replacing or recharging batteries when using such energy sources. Based on such study, a recommendation is to develop a dynamic system for energy management in real-time with the assessment of local renewable energy so that the system be reliable with minimum power consumption. Also, regarding energy management, we recommend providing backup energy sources (or using more than one energy source) or (a hybrid technique).

Abnormal event detection model using an improved ResNet101 in context aware surveillance system

AbstractSurveillance system plays a significant role for achieving security monitoring in the place of crowd areas. Offline monitoring of these crowd activity is quite challenging because it requires huge number of human resources for attaining efficient tracking. For shortcoming these issue automated and intelligent based system must be developed for efficiently monitor crowd and detect abnormal activity. However the existing methods faces issues like irrelevant features, high cost and process complexity. In this current research context aware surveillance‐system utilising hybrid ResNet101‐ANN is developed for effective abnormal activity detection. For this proposed approach video acquired from surveillance camera is considered as input. Then, acquired video is segmented into multiple frames. After that pre‐processing techniques such as denoising using mean filter, motion deblurring, contrast enhancement using Histogram Equalisation and canny edge detection is applied in this segmented frames. Further, the pre‐processed frame is fetched into hybrid ResNet101‐ANN classifier for abnormal event classification. Here, ResNet101 is used for extracting the features from the frames and Artificial neural network which replaces the fully connected layer of ResNet101 us used to detect the abnormal activity. If once abnormal‐events detected the context aware services generate alert to the user for preventing abnormal‐activities. Accuracy, precision, recall, and error values reached for the proposed‐model on simulation were 0.98, 0.98, 0.98 and 0.017 respectively. Using this proposed model effective crowd monitoring and abnormal activity detection can be achieved.

DeepVehicleSense: An Energy-Efficient Transportation Mode Recognition Leveraging Staged Deep Learning Over Sound Samples

In this paper, we present a new transportation mode recognition system for smartphones called DeepVehicleSense, which is widely applicable to mobile context-aware services. DeepVehicleSense aims at achieving three performance objectives: high accuracy, low latency, and low power consumption at once by exploiting sound characteristics captured from the built-in microphone while being on candidate transportations. To attain high energy efficiency, DeepVehicleSense adopts hierarchical accelerometer-based triggers that minimize the activation of the microphone of smartphones. Further, to achieve high accuracy and low latency, DeepVehicleSense makes use of non-linear filters that can best extract the transportation sound samples. For recognition of five different transportation modes, we design a deep learning based sound classifier using a novel deep neural network architecture with multiple branches. Our staged inference technique can significantly reduce runtime and energy consumption while maintaining high accuracy for the majority of samples. Through 263-hour datasets collected by seven different Android phone models, we demonstrate that DeepVehicleSense achieves the recognition accuracy of 97.44\% with only sound samples of 2 seconds at the power consumption of 35.08 mW on average for all-day monitoring.

Accurate and Reliable Service Recommendation Based on Bilateral Perception in Multi-Access Edge Computing

Multi-access edge computing (MEC) is an emerging computing paradigm that brings services from the centralized cloud to nearby network edge to improve users’ Quality of Experience (QoE). As massive services with dynamic Quality of Service (QoS) are available in MEC, it becomes challenging for users to find reliable services that satisfy their needs. Therefore, service recommendation technology is urgently needed in MEC. Although existing service recommendation methods work well on recommending popular services that users might be interested in, they fail to recommend services with reliable QoS in the MEC environment. To tackle this issue, an accurate and reliable service recommendation (ARSR) approach based on bilateral perception is proposed, which aims to proactively recommend reliable services by perceiving both users’ service demands and multi-QoS of candidate services. ARSR consists of three main steps. First, a user's service demand is estimated by a context-aware service demand prediction method based on an improved online deep learning model. Then, multiple QoS attributes of candidate services are forecasted by a multidimensional contexts-aware QoS prediction method based on an improved multi-task deep neural network. Finally, the optimal service is recommended to the user based on the predicted QoS. Extensive experiments have been carried out to verify the proposed approach and to prove its performance superiority.

At the Confluence of Artificial Intelligence and Edge Computing in IoT-Based Applications: A Review and New Perspectives.

Given its advantages in low latency, fast response, context-aware services, mobility, and privacy preservation, edge computing has emerged as the key support for intelligent applications and 5G/6G Internet of things (IoT) networks. This technology extends the cloud by providing intermediate services at the edge of the network and improving the quality of service for latency-sensitive applications. Many AI-based solutions with machine learning, deep learning, and swarm intelligence have exhibited the high potential to perform intelligent cognitive sensing, intelligent network management, big data analytics, and security enhancement for edge-based smart applications. Despite its many benefits, there are still concerns about the required capabilities of intelligent edge computing to deal with the computational complexity of machine learning techniques for big IoT data analytics. Resource constraints of edge computing, distributed computing, efficient orchestration, and synchronization of resources are all factors that require attention for quality of service improvement and cost-effective development of edge-based smart applications. In this context, this paper aims to explore the confluence of AI and edge in many application domains in order to leverage the potential of the existing research around these factors and identify new perspectives. The confluence of edge computing and AI improves the quality of user experience in emergency situations, such as in the Internet of vehicles, where critical inaccuracies or delays can lead to damage and accidents. These are the same factors that most studies have used to evaluate the success of an edge-based application. In this review, we first provide an in-depth analysis of the state of the art of AI in edge-based applications with a focus on eight application areas: smart agriculture, smart environment, smart grid, smart healthcare, smart industry, smart education, smart transportation, and security and privacy. Then, we present a qualitative comparison that emphasizes the main objective of the confluence, the roles and the use of artificial intelligence at the network edge, and the key enabling technologies for edge analytics. Then, open challenges, future research directions, and perspectives are identified and discussed. Finally, some conclusions are drawn.

Characterization and Prediction of Mobile Tasks

Mobile devices have become an increasingly ubiquitous part of our everyday life. We use mobile services to perform a broad range of tasks (e.g., booking travel or conducting remote office work), leading to often lengthy interactions with several distinct apps and services. Existing mobile systems handle mostly simple user needs, where a single app is taken as the unit of interaction. To understand users’ expectations and to provide context-aware services, it is important to model users’ interactions with their performed task in mind. To provide a comprehensive picture of common mobile tasks, we first conduct a small-scale user study to understand annotated mobile tasks in-depth, while we demonstrate that by using a set of features (temporal, similarity, and log sequence), we can identify if a pair of app usage belong to the same task effectively. Secondly, the proposed best task detection model is applied to a large-scale data set of commercial mobile app usage logs to infer characteristics of complex (multi-app) mobile tasks in the wild. By applying an unsupervised learning framework, we discover common mobile task types that span multiple apps based on various extracted characteristics. We observe that users generally perform 17 common tasks with 47 sub-tasks, ranging from “social media browsing” to “dining out” and “family entertainments”. Finally, we demonstrate that we can predict the next complex mobile task that users are likely to perform by leveraging features from the historically inferred mobile tasks and user contexts. Our work facilitates an in-depth understanding of mobile tasks at scale, enabling applications for promoting task-aware services.

Towards a Rule Modeling Framework for Context-aware Smart Service Systems

Since business rules aim at enforcing regulations in an organization, they are critical in governing business activities from a managerial standpoint. On the other hand, another type of rules has emerged in context-aware service systems: context rules. Context rules are employed for context reasoning to recommend and operate the right services in an appropriate manner. In this sense, context rules ensure the smartness of services in smart service systems. For decades, researchers and practitioners have addressed rule modelling and rule management in information systems and business services. However, in relation to context-aware services in smart service systems, there is a lack of exploring the rule aspect, especially considering how business rules and context rules are involved in such a system. The purpose of this paper is to propose a rule modelling framework (called RuCBS framework) for expressing rules in context-aware smart service systems over the three aspects of service science (Management, Science, and Engineering). The framework presents concepts, a meta-model that connects these concepts, and rule patterns. The framework is validated with a case study on banking services. Future research directions on rules in context-aware smart service systems are also discussed.

Machine Learning-Based Integrated Wireless Sensing and Positioning for Cellular Network

Integrated sensing and communication (ISAC) is one of the most promising technical directions in B5G/6G era. Nevertheless, it is challenging for network operators to provide proper human-centric services (HCSs), appropriately with the explosive growth of mobile terminals in constantly changing scenarios. Scenario sensing is the foundation to provide users with intelligent context-aware services, while the combination of machine learning (ML) and automatic data collection enables network operators to become more flexible in decision-making and planning. In this article, an ML-based integrated indoor/outdoor (IO) sensing and positioning framework is proposed to improve the potential of ML in the context of cellular network service management. First, large amounts of measurement reports (MRs) are collected over Layer 3 at user equipment (UE) and evolved NodeBs (eNBs) in urban areas, including both indoor and outdoor samples to simulate the minimization of drive test (MDT). Then, a random-forest-based IO classifier is implemented to sense the mobile scenarios and filter the positioning fingerprints. Subsequently, the weighted <inline-formula xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink"> <tex-math notation="LaTeX">$K$ </tex-math></inline-formula> -nearest neighbor (WKNN)-based Enhanced Cell ID (ECID) merging with the MR method is used for cellular positioning. The performance evaluation has shown that the positioning error of the MRs after denoising is effectively reduced compared with the conventional fingerprint-based positioning. Specifically, under the 67% standard defined by Federal Communications Commission (FCC), the resulting positioning error is about 4% lower than the case without preprocessing. In addition, comparative experiments are conducted to discuss the impact of different <inline-formula xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink"> <tex-math notation="LaTeX">$K$ </tex-math></inline-formula> values at different data sizes of fingerprints. Moreover, the further data analysis shows that MRs collected in the mild indoor (MI) scenarios have a positive effect on the overall positioning error, which indicates that refined contextualizations contribute to the cellular positioning for HCS upgrade.

Beyond AI-powered context-aware services: the role of human–AI collaboration

PurposeArtificial intelligence (AI) has gained significant momentum in recent years. Among AI-infused systems, one prominent application is context-aware systems. Although the fusion of AI and context awareness has given birth to personalized and timely AI-powered context-aware systems, several challenges still remain. Given the “black box” nature of AI, the authors propose that human–AI collaboration is essential for AI-powered context-aware services to eliminate uncertainty and evolve. To this end, this study aims to advance a research agenda for facilitators and outcomes of human–AI collaboration in AI-powered context-aware services.Design/methodology/approachSynthesizing the extant literature on AI and context awareness, the authors advance a theoretical framework that not only differentiates among the three phases of AI-powered context-aware services (i.e. context acquisition, context interpretation and context application) but also outlines plausible research directions for each stage.FindingsThe authors delve into the role of human–AI collaboration and derive future research questions from two directions, namely, the effects of AI-powered context-aware services design on human–AI collaboration and the impact of human–AI collaboration.Originality/valueThis study contributes to the extant literature by identifying knowledge gaps in human–AI collaboration for AI-powered context-aware services and putting forth research directions accordingly. In turn, their proposed framework yields actionable guidance for AI-powered context-aware service designers and practitioners.

Rapid prototyping and performance evaluation of ETSI MEC-based applications

The Multi-access Edge Computing (MEC) standard of the European Telecommunications Standards Institute (ETSI) will enable context-aware services for users of mobile 4G/5G networks. ETSI MEC application developers need tools to aid the design and the performance evaluation of their apps. During the early stages of deployment, they should be able to evaluate the performance impact of design choices - e.g., what round-trip delay can be expected due to the interplay of computation, communication and service consumption. When a prototype of the app exists, it needs to be tested live, under controllable conditions, to measure key performance indicators. In this paper, we present an open-source framework that allows developers to do all the above. Our framework is based on Simu5G, the OMNeT++ based simulator of 5G (NewRadio) and 4G (LTE) mobile networks. It includes models of ETSI MEC entities (i.e., MEC orchestrator, MEC host, etc.) and provides a standard-compliant RESTful interface towards application endpoints. Moreover, it can interface with external applications, and can also run in real time. Therefore, one can use it as a cradle to run a MEC app live, having underneath both 4G/5G data packet transport and MEC services based on information generated by the underlying emulated radio access network. We describe our framework and present a use-case of an emulated MEC-enabled 5G scenario.

An adaptive replica configuration mechanism based on predictive file popularity and queue balance in mobile edge computing environment

In the current internet of things era, various devices can provide more services by connecting to the Internet. However, the explosive growth of connected devices will cause cloud core overload and significant network delays. To overcome these overload and delay problems, the mobile edge computing (MEC) network is proposed to provide most of the computing and storage near the radio access network to reduce the traffic of the core cloud network and provide lower latency for the terminal. Mobile edge computing can work with third parties to develop multiple services, such as mobile big data analysis and context-aware services. However, while using the service, it may also encounter a large amount of popular data being accessed in a short period of time. Without proper handling of replica generation and deployment, even in low-latency environments, it can still kill the benefits of MEC due to increased access time. Although many scholars have proposed related issues for copy replication, there are still.parts that can be improved. To avoid the situation of insufficient availability of replicas, replica replication is performed, but infinite replicas may lead to a significant increase in traffic and waste of resources. And when deploying replicas, it is necessary to avoid placing them on congested nodes and consider how to achieve better load balancing. To improve the above problems, we synthesize the advantages of previous algorithms, make up for the shortcomings, and propose an adaptive replica configuration mechanism to predict the popularity of files and replicate replicas to low-blocking nodes. This method spreads the subsequent access workload by copying the popular file in advance to improve the system's overall performance.

CACS: A Context-Aware and Anonymous Communication Framework for an Enterprise Network Using SDN

The emergence of software-defined networking (SDN) has revolutionized the management of an enterprise network. The SDN-based design provides flexibility in network management, which spans over multiple applications, e.g., routing, switching, forwarding, and controlling. It reduces the reliance on vendor-specific devices and middlebox solutions, such as firewalls, intrusion detection systems (IDSs), intrusion prevention systems (IPSs), etc. Furthermore, due to the integration of different technologies, privacy is one of the core issues faced by the enterprise. Host anonymity is one of the techniques to safeguard against privacy attacks; however, the existing anonymization solutions provide better anonymity, but at the cost of higher latency and are most suited for Internet traffic. To tackle this issue in an enterprise network, we propose an SDN-based communication framework using enterprise integration patterns (EIPs) that offers anonymous communication in an enterprise environment. Host anonymity is achieved by replacing the real IP address with the spoofed IP address during the transmission of data packets inside the network. Unlike the traditional networks, SDN can modify the header fields of packets as they traverse in the network from the source to the destination. In addition to the host anonymity, this framework also provides context-aware communication by leveraging the SDN global visibility characteristic, where application services are discoverable on the network without disclosing the addresses of the application servers. Moreover, context-aware services enable network traffic to be routed based on the application-layer services rather than the network-layer information. In the end, evaluation of the proposed framework is carried out with respect to the performance of anonymous communication, computational complexity, and security of the complete proposed framework. In addition, we also highlighted that the proposed framework is more suitable for heterogeneous network environments such as Internet of Things-based solutions.

A Two-Phase Machine Learning Framework for Context-Aware Service Selection to Empower People with Disabilities.

The use of software and IoT services is increasing significantly among people with special needs, who constitute 15% of the world’s population. However, selecting appropriate services to create a composite assistive service based on the evolving needs and context of disabled user groups remains a challenging research endeavor. Our research applies a scenario-based design technique to contribute (1) an inclusive disability ontology for assistive service selection, (2) semi-synthetic generated disability service datasets, and (3) a machine learning (ML) framework to choose services adaptively to suit the dynamic requirements of people with special needs. The ML-based selection framework is applied in two complementary phases. In the first phase, all available atomic tasks are assessed to determine their appropriateness to the user goal and profiles, whereas in the subsequent phase, the list of service providers is narrowed by matching their quality-of-service factors against the context and characteristics of the disabled person. Our methodology is centered around a myriad of user characteristics, including their disability profile, preferences, environment, and available IT resources. To this end, we extended the widely used QWS V2.0 and WS-DREAM web services datasets with a fusion of selected accessibility features. To ascertain the validity of our approach, we compared its performance against common multi-criteria decision making (MCDM) models, namely AHP, SAW, PROMETHEE, and TOPSIS. The findings demonstrate superior service selection accuracy in contrast to the other methods while ensuring accessibility requirements are satisfied.

Design and Implementation of Online Meticulous Painting System Based on Mobile IoT Service.

Contemporary art is gradually diversifying, and the aesthetics of artists have changed and progressed together with the times. Some brush painters, while inheriting the essence of tradition, continue to summarize and search for a painting language in line with the atmosphere of the times, gradually shifting their perspective to a separate, tiny “corner” of life, recording the painter's own subtle emotional fluctuations through such “tiny” scenes. The approach models services through a context-aware service model and establishes flexible driving relationships between IoT data and services through context and scenario concepts. Moreover, it reduces the amount of business data to be processed by the system through contextual events. The process of the approach for service development is described, and relevant context-aware service development tools are introduced. Finally, the feasibility of the approach proposed in this article is illustrated by a service development example. An event-triggered sensor node clustering algorithm based on beamforming data transmission technology is proposed, which can improve network energy efficiency and clustering success rate. Meanwhile, a clustering algorithm based on the energy consumption rate of sensor nodes is designed to optimize the task allocation and scheduling of mobile node access, reduce the node charging delay, and enable the sensor nodes in the network to get energy replenishment in a timely and reasonable manner according to their own energy consumption rate demand.

Energy-Aware Edge Association for Cluster-Based Personalized Federated Learning

Federated Learning (FL) over wireless network enables data-conscious services by leveraging the ubiquitous intelligence at network edge for privacy-preserving model training. As the proliferation of context-aware services, the diversified personal preferences cause disagreeing conditional distributions among user data, which leads to poor inference performance. In this sense, clustered federated learning is proposed to group user devices with similar preference and provide each cluster with a personalized model. This calls for innovative design in edge association that involves user clustering and also resource management optimization. We formulate an accuracy-cost trade-off optimization problem by jointly considering model accuracy, communication resource allocation and energy consumption. To comply with parameter encryption techniques in FL, we propose an iterative solution procedure which employs deep reinforcement learning based approach at cloud server for edge association. The reward function consists of minimized energy consumption at each base station and the averaged model accuracy of all users. Under our proposed solution, multiple edge base station are fully exploited to realize cost efficient personalized federated learning without any prior knowledge on model parameters. Simulation results show that our proposed strategy outperforms existing strategies in achieving accurate learning at low energy cost.