arXiv subjects

Xincao Xu

Publications and source records attributed to Xincao Xu.

8 recordsLinked to original sources

arXiv2026

FedCoE: Bridging Generalization and Personalization via Federated Coordinated Dual-level MoEs

Federated Learning (FL) has emerged as a promising paradigm for privacy-preserving distributed learning. However, existing FL methods face a fundamental challenge. Traditional averaging-based approaches suffer from parameter divergence under non-IID conditions, while personalized FL methods overfit to local data and fail to generalize to new clients (cold-start problem). Mixture-of-Experts naturally addresses this by routing heterogeneous data to specialized experts rather than forcing uniform aggregation. In this paper, we propose FedCoE, a Federated Coordinated dual-level mixture-of-Experts framework that effectively balances global generalization with local personalization. FedCoE maintains multiple independent global expert models on the server and employs a shared gating network to dynamically model client-expert correlations during aggregation, effectively mitigating expert drift and gating inconsistency. To address the cold-start challenge, we introduce an adaptive mechanism that enables new clients to immediately leverage the global expert pool without extensive local training. Extensive experiments demonstrate that FedCoE achieves 78.00% global accuracy and 89.32% personalized accuracy on average, outperforming the baseline by 8.82% and 29.19%, respectively. In cold-start scenarios, FedCoE delivers 77.27% accuracy without any local fine-tuning, outperforming baselines by over 12.54%.

cs.LG↗

arXiv2026

AutoMCU: Feasibility-First MCU Neural Network Customization via LLM-based Multi-Agent Systems

Deploying neural networks on microcontroller units (MCUs) is critical for edge intelligence but remains challenging due to tight memory, storage, and computation constraints. Existing approaches, such as model compression and hardware-aware neural architecture search (HW-NAS), often depend on proxy metrics, incur high search cost, and do not fully bridge the gap between architecture design and verified deployment. This paper presents AutoMCU, a feasibility-first large language model (LLM)-based multi-agent system for automated neural network customization under MCU constraints. Given natural-language task requirements and hardware specifications, AutoMCU iteratively generates structured architecture candidates, filters infeasible designs through vendor toolchain feedback before training, evaluates feasible models under a controlled protocol, and verifies deployability through backend-grounded deployment analysis. AutoMCU includes two key mechanisms: 1) hardware-in-the-loop architecture generation for early elimination of undeployable candidates under RAM and Flash constraints, and 2) state-isolated multi-agent scheduling for stable coordination of proposal, training, evaluation, and deployment stages. Experiments on CIFAR-10 and CIFAR-100 under strict MCU constraints show that AutoMCU achieves competitive accuracy while reducing customization time to about 1--2 hours, compared with hundreds of GPU hours for representative MCU-oriented HW-NAS baselines. Comparisons with ColabNAS and the LLM-based NAS method GENIUS on NAS-Bench-201 further demonstrate the effectiveness and stability of AutoMCU. Real-device deployments on multiple STM32 microcontrollers validate its practical applicability to MCU-scale edge intelligence.

cs.LG↗

arXiv2023

Quality-Cost Trade-off on Constructing Logical Views for Vehicular Cyber-Physical Systems: A Deep Reinforcement Learning Approach

With the development of sensing technologies, vehicle-to-everything (V2X) communications, edge computing paradigm, vehicular cyber-physical systems (VCPS) are emerging as the most fundamental platform for realizing future intelligent transportation systems (ITSs). In particular, the construction of logical views at the edge nodes based on heterogeneous information sensing and uploading are critical to the realization of VCPS. However, a higher-quality view in terms of timeliness and accuracy may require higher cost on sensing and uploading. In view of this, this paper is dedicated to striking a balance between the quality and the cost for constructing logical views of VCPS. Specifically, we first derive an information sensing model based on multi-class M/G/1 priority queue and a data uploading model based on reliability-guaranteed vehicle-to-infrastructure (V2I) communications. On this basis, we design two metrics, namely, age of view (AoV) and cost of view (CoV), simultaneously. Then, we formulate a bi-objective problem to maximize the AoV and minimize the CoV. Further, we propose a distributed distributional deep deterministic policy gradient (D4PG) solution to determine sensing information, frequency, uploading priority, transmission power, and V2I bandwidth. Finally, we build a simulation model and give a comprehensive performance evaluation, and the simulation results conclusively demonstrate the superiority of the proposed solution.

cs.NI↗

arXiv2023

Cooperative Sensing and Heterogeneous Information Fusion in VCPS: A Multi-agent Deep Reinforcement Learning Approach

Cooperative sensing and heterogeneous information fusion are critical to realize vehicular cyber-physical systems (VCPSs). This paper makes the first attempt to quantitatively measure the quality of VCPS by designing a new metric called Age of View (AoV). Specifically, we first present the system architecture where heterogeneous information can be cooperatively sensed and uploaded via vehicle-to-infrastructure (V2I) communications in vehicular edge computing (VEC). Logical views are constructed by fusing the heterogeneous information at edge nodes. Further, we formulate the problem by deriving a cooperative sensing model based on the multi-class M/G/1 priority queue, and defining the AoV by modeling the timeliness, completeness and consistency of the logical views. On this basis, a multi-agent deep reinforcement learning solution is proposed. In particular, the system state includes vehicle sensed information, edge cached information and view requirements. The vehicle action space consists of the sensing frequencies and uploading priorities of information. A difference-reward-based credit assignment is designed to divide the system reward, which is defined as the VCPS quality, into the difference reward for vehicles. Edge node allocates V2I bandwidth to vehicles based on predicted vehicle trajectories and view requirements. Finally, we build the simulation model and give a comprehensive performance evaluation, which conclusively demonstrates the superiority of the proposed solution.

cs.NI↗

arXiv2023

Cooperative Sensing and Uploading for Quality-Cost Tradeoff of Digital Twins in VEC

Recent advances in sensing technologies, wireless communications, and computing paradigms drive the evolution of vehicles in becoming an intelligent and electronic consumer products. This paper investigates enabling digital twins in vehicular edge computing (DT-VEC) via cooperative sensing and uploading, and makes the first attempt to achieve the quality-cost tradeoff in DT-VEC. First, a DT-VEC architecture is presented, where the heterogeneous information can be sensed by vehicles and uploaded to the edge node via vehicle-to-infrastructure (V2I) communications. The digital twins are modeled based on the sensed information, which are utilized to from the logical view to reflect the real-time status of the physical vehicular environment. Second, we derive the cooperative sensing model and the V2I uploading model by considering the timeliness and consistency of digital twins, and the redundancy, sensing cost, and transmission cost. On this basis, a bi-objective problem is formulated to maximize the system quality and minimize the system cost. Third, we propose a solution based on multi-agent multi-objective (MAMO) deep reinforcement learning, where a dueling critic network is proposed to evaluate the agent action based on the value of state and the advantage of action. Finally, we give a comprehensive performance evaluation, demonstrating the superiority of MAMO.

cs.NI↗

arXiv2022

Joint Task Offloading and Resource Optimization in NOMA-based Vehicular Edge Computing: A Game-Theoretic DRL Approach

Vehicular edge computing (VEC) becomes a promising paradigm for the development of emerging intelligent transportation systems. Nevertheless, the limited resources and massive transmission demands bring great challenges on implementing vehicular applications with stringent deadline requirements. This work presents a non-orthogonal multiple access (NOMA) based architecture in VEC, where heterogeneous edge nodes are cooperated for real-time task processing. We derive a vehicle-to-infrastructure (V2I) transmission model by considering both intra-edge and inter-edge interferences and formulate a cooperative resource optimization (CRO) problem by jointly optimizing the task offloading and resource allocation, aiming at maximizing the service ratio. Further, we decompose the CRO into two subproblems, namely, task offloading and resource allocation. In particular, the task offloading subproblem is modeled as an exact potential game (EPG), and a multi-agent distributed distributional deep deterministic policy gradient (MAD4PG) is proposed to achieve the Nash equilibrium. The resource allocation subproblem is divided into two independent convex optimization problems, and an optimal solution is proposed by using a gradient-based iterative method and KKT condition. Finally, we build the simulation model based on real-world vehicle trajectories and give a comprehensive performance evaluation, which conclusively demonstrates the superiority of the proposed solutions.

cs.NI↗

arXiv2022

Age of View: A New Metric for Evaluating Heterogeneous Information Fusion in Vehicular Cyber-Physical Systems

Heterogeneous information fusion is one of the most critical issues for realizing vehicular cyber-physical systems (VCPSs). This work makes the first attempt at quantitatively measuring the quality of heterogeneous information fusion in VCPS by designing a new metric called Age of View (AoV). Specifically, we derive a sensing model based on a multi-class M/G/1 priority queue and a transmission model based on Shannon theory. On this basis, we formally define AoV by modeling the timeliness, completeness, and consistency of the heterogeneous information fusion in VCPS and formulate the problem aiming to minimize the system's average AoV. Further, we propose a new solution called Multi-agent Difference-Reward-based deep reinforcement learning with a Greedy Bandwidth Allocation (MDR-GBA) to solve the problem. In particular, each vehicle acts as an independent agent and decides the sensing frequencies and uploading priorities of heterogeneous information. Meanwhile, the roadside unit (RSU) decides the Vehicle-to-Infrastructure (V2I) bandwidth allocation for each vehicle based on a greedy scheme. Finally, we build the simulation model and compare the performance of the proposed solution with state-of-the-art algorithms. The experimental results conclusively demonstrate the significance of the new metric and the superiority of the proposed solution.

cs.NI↗

arXiv2020

Vehicular Fog Computing Enabled Real-time Collision Warning via Trajectory Calibration

Vehicular fog computing (VFC) has been envisioned as a promising paradigm for enabling a variety of emerging intelligent transportation systems (ITS). However, due to inevitable as well as non-negligible issues in wireless communication, including transmission latency and packet loss, it is still challenging in implementing safety-critical applications, such as real-time collision warning in vehicular networks. In this paper, we present a vehicular fog computing architecture, aiming at supporting effective and real-time collision warning by offloading computation and communication overheads to distributed fog nodes. With the system architecture, we further propose a trajectory calibration based collision warning (TCCW) algorithm along with tailored communication protocols. Specifically, an application-layer vehicular-to-infrastructure (V2I) communication delay is fitted by the Stable distribution with real-world field testing data. Then, a packet loss detection mechanism is designed. Finally, TCCW calibrates real-time vehicle trajectories based on received vehicle status including GPS coordinates, velocity, acceleration, heading direction, as well as the estimation of communication delay and the detection of packet loss. For performance evaluation, we build the simulation model and implement conventional solutions including cloud-based warning and fog-based warning without calibration for comparison. Real-vehicle trajectories are extracted as the input, and the simulation results demonstrate that the effectiveness of TCCW in terms of the highest precision and recall in a wide range of scenarios.

cs.NI↗