Autonomous navigation for Unmanned Aerial Vehicles (UAVs) presents significant challenges due to the limited onboard computational resources, which often restrict deployed deep neural networks to shallow architectures incapable of handling complex environments. Additionally, offloading tasks to remote edge servers introduces high latency, creating an inherent trade-off in system design. To address these limitations, we propose CoDrone—the first cloud-edge-end collaborative computing framework that integrates foundation models into autonomous UAV cruising scenarios—effectively leveraging foundation models to enhance the performance of resource-constrained unmanned aerial vehicle platforms. To reduce both onboard computation and data transmission overhead, CoDrone employs grayscale imagery for the navigation model. When enhanced environmental perception is required, CoDrone leverages the edge-assisted foundation model Depth Anything V2 for depth estimation and introduces a novel, one-dimensional occupancy grid–based navigation method—enabling fine-grained scene understanding while significantly advancing the efficiency and representational simplicity of autonomous navigation. A key component of CoDrone is a Deep Reinforcement Learning (DRL)-based neural scheduler that seamlessly integrates depth estimation with autonomous navigation decisions, enabling real-time adaptation to dynamic environments. Furthermore, the framework introduces a UAV-specific vision language interaction module, which incorporates domain-tailored low-level flight primitives to enable effective interaction between the cloud foundation model, the Vision Language model, and the UAV. The introduction of VLM enhances open-set reasoning capabilities in complex and previously unseen scenarios. We implement a prototype of CoDrone and conduct extensive evaluations in the AirSim simulation environment. Experimental results demonstrate that CoDrone significantly outperforms baseline methods under varying flight speeds and network conditions, achieving a 40% increase in average flight distance and a 5% improvement in average Quality of Navigation.

Benefiting from hardware upgrades and deep learning techniques, more and more end devices can independently support a variety of intelligent applications. Further powered by edge computing technologies, the end-edge collaboration paradigm becomes one mainstream approach for achieving advanced edge intelligence (EI). To fully exploit the system resources, it is desirable to coordinate diverse EI services efficiently. Thus, we present a novel framework to jointly optimize the cost-performance trade-off for two distinct but typical EI services, where end devices simultaneously perform federated learning (FL) model training and conduct model inference with the assistance of edge offloading. However, balancing the long-term cost-performance trade-off is highly non-trivial, especially in the absence of knowledge of future system dynamics. Moreover, the capacity heterogeneity further increases the difficulty of service coordination among resource-limited end devices. To overcome these challenges, we first analyze the optimality of inference offloading decisions with and without FL model training and quantify their mutual effects due to local resource contention. By incorporating the loss estimation of FL training model, we then propose a novel proactive policy with theoretical guarantees, which proactively controls the stopping of FL training procedure to balance well the trade-offs between FL model performance and resource costs while fulfilling the inference performance requirements. Extensive results show the efficiency and robustness of our proposed algorithm for EI service coordination in dynamic end-edge collaboration scenarios.

3D Gaussian Splatting (3DGS) has recently emerged as a prominent technique in novel view synthesis. The superior performance of 3DGS has catalyzed an increasing number of 3DGS- based applications in edge scenarios, where 3DGS is utilized for various purposes, such as scene representation, comprehension, and generation. Meanwhile, these edge applications also serve as primary sources of scene observations for producing 3DGS models. However, the intensive computation involved in 3DGS training and the massive number of 3D Gaussian primitives required for high-resolution scene repre-sentation hinder the effectiveness of in-situ 3DGS training on off-the-shelf edge devices, whether using standalone training or Data-Distributed-Parallel (DDP) training. To address this issue, this work proposes MIX3D, a novel mixed representation for communication-efficient distributed 3DGS training in edge scenarios. MIX3D features a global sparse sub-model and various local dense sub-models, where the sparse sub-model encodes coarse-grained appearance for the entire scene, and each dense sub-model targets fine-grained details for a specific region of the scene. Extensive evaluations on a four-device edge cluster demonstrate the effectiveness of our developed distributed 3DGS training workflow based on MIX3D, achieving reductions in training time up to 86.6% compared to vanilla DDP training and an average speedup of 3.767x over standalone training.

Graph Neural Networks (GNNs) have gained growing interest in miscellaneous applications owing to their outstanding ability in extracting latent representation on graph structures. To render GNN-based service for IoT-driven smart applications, traditional model serving paradigms usually resort to the cloud by fully uploading geo-distributed input data to remote datacenters. However, our empirical measurements reveal the significant communication overhead of such cloud-based serving and highlight the profound potential in applying the emerging fog computing. To maximize the architectural benefits brought by fog computing, in this paper, we present Fograph, a novel distributed real-time GNN inference framework that leverages diverse and dynamic resources of multiple fog nodes in proximity to IoT data sources. By introducing heterogeneity-aware execution planning and GNN-specific compression techniques, Fograph tailors its design to well accommodate the unique characteristics of GNN serving in fog environments. Prototype-based evaluation and case study demonstrate that Fograph significantly outperforms the state-of-the-art cloud serving and fog deployment by up to 5.39$\times$execution speedup and 6.84$\times$throughput improvement.

To identify dense and small-size pedestrians in surveillance systems, high-resolution cameras are widely deployed, where high-resolution images are captured and delivered to off-the-shelf pedestrian detection models. However, given the highly computation-intensive workload brought by the high resolution, the resource-constrained cameras fail to afford accurate inference in real time. To address that, we propose Hode, an offloaded video analytic framework that utilizes multiple edge nodes in proximity to expedite pedestrian detection with high-resolution inputs. Specifically, Hode can intelligently split high-resolution images into respective regions and then offload them to distributed edge nodes to perform pedestrian detection in parallel. A spatio-temporal flow filtering method is designed to enable context-aware region partitioning, as well as a DRL-based scheduling algorithm to allow accuracy-aware load balance among heterogeneous edge nodes. Extensive evaluation results using realistic prototypes show that Hode can achieve up to 2.01× speedup with very mild accuracy loss.

Despite the popularity of Serverless computing, there are insufficient efforts dedicated to Serverless workflows (i.e., Serverless function orchestration), particularly for Serverless edge computing. In this paper, we first identify the challenges of deploying the state-of-the-art cloud-oriented Serverless workflow scheduling on resource-constrained edge devices, then propose to model Serverless workflows with behavior trees, and finally reveal our key observations and preliminary results for behavior tree-based Serverless workflow scheduling.

Federated Learning (FL) has been a promising paradigm in distributed machine learning that enables in-situ model training and global model aggregation. While it can well preserve private data for end users, to apply it efficiently on IoT devices yet suffer from their inherent variants: their available computing resources are typically constrained, heterogeneous, and changing dynamically. Existing works deploy FL on IoT devices by pruning a sparse model or adopting a tiny counterpart, which alleviates the workload but may have negative impacts on model accuracy. To address these issues, we propose Eco-FL, a novel Edge Collaborative pipeline based Federated Learning framework. On the client side, each IoT device collaborates with trusted available devices in proximity to perform pipeline training, enabling local training acceleration with efficient augmented resource orchestration. On the server side, Eco-FL adopts a novel grouping-based hierarchical architecture that combines synchronous intra-group aggregation and asynchronous inter-group aggregation, where a heterogeneity-aware dynamic grouping strategy that jointly considers response latency and data distribution is developed. To tackle the resource fluctuation during the runtime, Eco-FL further applies an adaptive scheduling policy to judiciously adjust workload allocation and client grouping at different levels. Extensive experimental results using both prototype and simulation show that, compared to state-of-the-art methods, Eco-FL can upgrade the training accuracy by up to 26.3%, reduce the local training time by up to 61.5%, and improve the local training throughput by up to 2.6 ×.

Graph Neural Networks (GNNs) have gained growing interest in miscellaneous applications owing to their outstanding ability in extracting latent representation on graph structures. To render GNN-based service for IoT-driven smart applications, the traditional model serving paradigm resorts to the cloud by fully uploading the geo-distributed input data to the remote datacenter. However, our empirical measurements reveal the significant communication overhead of such cloud-based serving and highlight the profound potential in applying the emerging fog computing. To maximize the architectural benefits brought by fog computing, in this paper, we present Fograph, a novel distributed real-time GNN inference framework that leverages diverse resources of multiple fog nodes in proximity to IoT data sources. By introducing heterogeneity-aware execution planning and GNN-specific compression techniques, Fograph tailors its design to well accommodate the unique characteristics of GNN serving in fog environment. Prototype-based evaluation and case study demonstrate that Fograph significantly outperforms the state-of-the-art cloud serving and vanilla fog deployment by up to 5.39 × execution speedup and 6.84 × throughput improvement.

With the growing size of the cloud data center, the high scheduling efficiency over massive-scale cloud servers is hard to achieve, particularly when the scheduler requires the full real-time cloud resource information for decision making. Moreover, most data centers only run latency-critical online services, resulting in low resource utilization. To solve these problems, we propose a Compressive Sensing based Predictive Online Scheduling (CSPOS) algorithm. To mitigate the bottleneck of transferring massive resource information of all cloud servers to the scheduler, we propose to transfer sampled data from a small subset of servers to the scheduler and recover the full cloud resource information by compressive sensing. We then propose a predictive online learning algorithm that efficiently colocates the online services and batch jobs, in order to boost the resource utilization of the data center. Our experiments show that the CSPOS model achieves outstanding scheduling efficiency under various settings and is able to greatly increase the resource usage of a data center. We also illustrate that the running time of the CSPOS model is very small and has negligible effects on the scheduling system.

Recently, the explosion of resource-hungry and delay-sensitive Internet-of-Things (IoT) applications as exemplified by wearable appliances, video surveillance, and connected vehicles have posed great challenges on the underlying IoT devices which typically have limited computation resource. In response, computation offloading is envisioned as a promising approach to augmenting capability of IoT devices. Toward real-time and efficient computation offloading, in this paper we propose a novel edge resource pooling framework, in which a massive crowd of devices at the network edge exploit device-to-device (D2D) collaboration for pooling and sharing computation resource with each other. Specifically, we first formulate the utility maximization problem under both computation and communication constraints as a mixed-integer linear programming problem, which is further proven to be NP-hard. To address this challenge, we propose a greedy heuristic based on the classical maximum network flow problem, and thus to schedule the task offloading in a cost-efficient manner. By considering the case that a centralized controller (e.g., a network operator) is not available, a decentralized task offloading scheme is further proposed, in which IoT devices communicate and determine D2D offloading strategy locally. Rigorous theoretical analysis and extensive evaluations demonstrate the effectiveness of the proposed algorithms.

With the breakthroughs in deep learning, the recent years have witnessed a booming of artificial intelligence (AI) applications and services, spanning from personal assistant to recommendation systems to video/audio surveillance. More recently, with the proliferation of mobile computing and Internet of Things (IoT), billions of mobile and IoT devices are connected to the Internet, generating zillions bytes of data at the network edge. Driving by this trend, there is an urgent need to push the AI frontiers to the network edge so as to fully unleash the potential of the edge big data. To meet this demand, edge computing, an emerging paradigm that pushes computing tasks and services from the network core to the network edge, has been widely recognized as a promising solution. The resulted new interdiscipline, edge AI or edge intelligence (EI), is beginning to receive a tremendous amount of interest. However, research on EI is still in its infancy stage, and a dedicated venue for exchanging the recent advances of EI is highly desired by both the computer system and AI communities. To this end, we conduct a comprehensive survey of the recent research efforts on EI. Specifically, we first review the background and motivation for AI running at the network edge. We then provide an overview of the overarching architectures, frameworks, and emerging key technologies for deep learning model toward training/inference at the network edge. Finally, we discuss future research opportunities on EI. We believe that this survey will elicit escalating attentions, stimulate fruitful discussions, and inspire further research ideas on EI.

The explosion of resource-hungry mobile applications has posed great challenges on the underlying mobile devices which typically have limited computation resource. In response, device-to-device (D2D) computation offloading is envisioned as a promising approach to the problem by gearing resource-rich devices and resource-poor devices. Towards real-time and efficient computation offloading, in this paper, we proposed a novel edge resource pooling framework called ERP, in which a massive crowd of devices at the network edge exploit D2D collaboration for pooling and sharing computation resource with each other. Specifically, we first formulate the utility maximization problem under both computation and communication constraints as a mixed-integer linear programming (MILP), which is further proven to be NP-hard. To address this challenge, we propose a centralized greedy heuristic based on the classical maximum network flow problem, which schedules the task offloading in a cost-efficient manner. Rigorous theoretical analysis and extensive evaluations demonstrate the effectiveness of the heuristic to some extent.