Yaxin Li
Current Organizations / Institutes
The organizations/institutes from the scholar's latest article
Previous Organizations / Institutes
The organizations / institutes from the scholar's other articles in the past ten years
Research Subjects
Research Interests
Yaxin Li has a total of 27 co-authors. Here are the top co-authors by number of shared publications:
- Yu Wang
School of Electrical Engineering and Information, Southwest Petroleum University, Chengdu, China
11 shared publications - Cheng Ren
School of Electrical Engineering and Information, Southwest Petroleum University, Chengdu, China
10 shared publications Haiyun Xiang
Network and Information Center, Southwest Petroleum University, Chengdu, China
2 shared publications- Hao Li
School of Electrical Engineering and Information, Southwest Petroleum University, Chengdu, China
2 shared publications Jintao Hu
The Seventh Research Division and the Center for Information and Control, School of Automation Science and Electri-cal Engineering, Beihang University
2 shared publications- Tao Yang
Department of Neurosurgery, The General Hospital of Western Theater Command, Chengdu, China
2 shared publications
Latest Publications
- 7 April 2026
- Published by Institute of Electrical and Electronics Engineers (IEEE) in IEEE Transactions on Network and Service Management
- Vol. 23, 4135-4145
GCN-Transformer-Assisted Live SFC Migration With Hierarchical Reinforcement Learning in Mobile Edge Computing
Empowered by network function virtualization (NFV), mobile edge computing aims to provide low latency and ultra reliable network services to mobile end users, achieved as a service function chain (SFC) consisting of a series of ordered virtual network functions (VNFs). Due to user mobility, live SFC migration is imperative to avoid Quality of Service (QoS) degradation. Recent advances mainly make separate decisions on VNF node remapping and migration path routing in a heuristic manner, or implement both through reinforcement learning within a single agent of ill-defined policy and action space. In this paper, given next access node, we first formulate the live SFC migration problem as an integer linear programming (ILP) model to achieve optimal solutions. Then, we present HRL-QC, a hierarchical reinforcement learning framework that jointly optimizes VNF destination node remapping, migration path and post-migration service path selections for QoS-aware and cost-efficient live SFC migration. A GCN-Transformer block is introduced to capture long-range VNF-to-physical node dependencies, while a two-level actor-critic design couples the decision-makings through inter-level reward passing. Extensive evaluations show that HRL-QC outperforms the state-of-the-art in energy consumption, migration time, end-to-end service delay, and migration success rate, while remaining within a small margin of the optimal ILP solution.
- 7 April 2026
- Published by Institute of Electrical and Electronics Engineers (IEEE) in IEEE Transactions on Network and Service Management
- Vol. 23, 4135-4145
GCN-Transformer-Assisted Live SFC Migration With Hierarchical Reinforcement Learning in Mobile Edge Computing
Empowered by network function virtualization (NFV), mobile edge computing aims to provide low latency and ultra reliable network services to mobile end users, achieved as a service function chain (SFC) consisting of a series of ordered virtual network functions (VNFs). Due to user mobility, live SFC migration is imperative to avoid Quality of Service (QoS) degradation. Recent advances mainly make separate decisions on VNF node remapping and migration path routing in a heuristic manner, or implement both through reinforcement learning within a single agent of ill-defined policy and action space. In this paper, given next access node, we first formulate the live SFC migration problem as an integer linear programming (ILP) model to achieve optimal solutions. Then, we present HRL-QC, a hierarchical reinforcement learning framework that jointly optimizes VNF destination node remapping, migration path and post-migration service path selections for QoS-aware and cost-efficient live SFC migration. A GCN-Transformer block is introduced to capture long-range VNF-to-physical node dependencies, while a two-level actor-critic design couples the decision-makings through inter-level reward passing. Extensive evaluations show that HRL-QC outperforms the state-of-the-art in energy consumption, migration time, end-to-end service delay, and migration success rate, while remaining within a small margin of the optimal ILP solution.
- 26 November 2025
- Published by Elsevier in Ocean Engineering
- Vol. 344
Fixed-time prescribed performance control for trajectory tracking of underactuated AUVs under thrust saturation
- 26 November 2025
- Published by Elsevier in Ocean Engineering
- Vol. 344
Fixed-time prescribed performance control for trajectory tracking of underactuated AUVs under thrust saturation
- 3 February 2025
- Published by Institute of Electrical and Electronics Engineers (IEEE) in IEEE Transactions on Network and Service Management
- Vol. 22 (2) , 1801-1811
A Fastformer Assisted DRL Method on Energy Efficient and Interference Aware Service Provisioning
Network function virtualization (NFV) empowered by virtualization technology can achieve flexible virtual network function (VNF) placement. To improve resource utilization and energy efficiency, different VNFs tend to be co-located on common servers, which inevitably intrigues VNF performance degradation induced by hardware resource competition. The problem of energy-efficient and interference-aware service function chain (SFC) provisioning is considered in this paper and envisioned to yield minimum activated servers and maximum average throughput. It is formulated as a mixed integer linear programming (MILP) model to achieve optimal solutions. Then, a gale-shapley based offline approximation algorithm is designed through bipartite matching, to yield an SFC allocation decision in one go with proved competitive ratio. In online scenario, Transformer and its efficient model Fastformer, combined with Graph Attention Network (GAT) respectively, are introduced into deep reinforcement learning (DRL) structure for the first time to quickly and accurately abstract features of substrate network and SFC. A DRL-based Fastformer-assisted energy efficient and interference aware SFC provisioning (DRL-EI) algorithm is proposed with an elaborately designed reward function to balance energy consumption and VNF interference. Simulations indicate the gap between DRL-EI and MILP is marginal. DRL-EI outperforms state-of-art work in terms of energy consumption, VNF normalized throughput and acceptance rate.
- 3 February 2025
- Published by Institute of Electrical and Electronics Engineers (IEEE) in IEEE Transactions on Network and Service Management
- Vol. 22 (2) , 1801-1811
A Fastformer Assisted DRL Method on Energy Efficient and Interference Aware Service Provisioning
Network function virtualization (NFV) empowered by virtualization technology can achieve flexible virtual network function (VNF) placement. To improve resource utilization and energy efficiency, different VNFs tend to be co-located on common servers, which inevitably intrigues VNF performance degradation induced by hardware resource competition. The problem of energy-efficient and interference-aware service function chain (SFC) provisioning is considered in this paper and envisioned to yield minimum activated servers and maximum average throughput. It is formulated as a mixed integer linear programming (MILP) model to achieve optimal solutions. Then, a gale-shapley based offline approximation algorithm is designed through bipartite matching, to yield an SFC allocation decision in one go with proved competitive ratio. In online scenario, Transformer and its efficient model Fastformer, combined with Graph Attention Network (GAT) respectively, are introduced into deep reinforcement learning (DRL) structure for the first time to quickly and accurately abstract features of substrate network and SFC. A DRL-based Fastformer-assisted energy efficient and interference aware SFC provisioning (DRL-EI) algorithm is proposed with an elaborately designed reward function to balance energy consumption and VNF interference. Simulations indicate the gap between DRL-EI and MILP is marginal. DRL-EI outperforms state-of-art work in terms of energy consumption, VNF normalized throughput and acceptance rate.
- 24 April 2024
- Published by Institute of Electrical and Electronics Engineers (IEEE) in IEEE Transactions on Network and Service Management
- Vol. 21 (4) , 4531-4542
On Efficient VNF-FG Design in IoT Networks
In recent times, it has been witnessed that an increasing number of Internet connected devices impose an huge challenge on Internet of Thing (IoT) networks, which provides diverse and complex network services through edge computing empowered IoT terminals. A network service can be formally represented by Virtualized Network Function Forwarding Graph (VNF-FG) with the advent of Network Function Virtualization (NFV) technology. Previous researches mainly focus on VNF-FG embedding (VNF-FGE) and take VNF-FG as the input. In this paper, we investigate the design of VNF-FG, which is required to be a Directed Acyclic Graph (DAG) and achieved by the IoT terminal, in two scenarios. In static scenario, for a set of traffic flows arriving at the IoT terminal and requesting different network services, an ILP model Ps including loop prevention constraints is well formulated. An approximation algorithm AFGC with competitive ratio O(1+(|R|-1)α), α (0, 1) is then designed, which thoroughly search all key instances to run comprehensive loop break. In dynamic scenario for a flow request on the fly reaching an IoT terminal, an ILP model Pd and another approximation algorithm DFGU with a competitive ratio O(1 + |SCr||Fr| ) are developed, giving priority to reuse of existing topology to generate an augmented VNF-FG of minimum size. Simulation results indicate AFGC and DFGU outperform state-of-art work, and the gap between each of the two algorithms and their respective ILP models is marginal.
- 24 April 2024
- Published by Institute of Electrical and Electronics Engineers (IEEE) in IEEE Transactions on Network and Service Management
- Vol. 21 (4) , 4531-4542
On Efficient VNF-FG Design in IoT Networks
In recent times, it has been witnessed that an increasing number of Internet connected devices impose an huge challenge on Internet of Thing (IoT) networks, which provides diverse and complex network services through edge computing empowered IoT terminals. A network service can be formally represented by Virtualized Network Function Forwarding Graph (VNF-FG) with the advent of Network Function Virtualization (NFV) technology. Previous researches mainly focus on VNF-FG embedding (VNF-FGE) and take VNF-FG as the input. In this paper, we investigate the design of VNF-FG, which is required to be a Directed Acyclic Graph (DAG) and achieved by the IoT terminal, in two scenarios. In static scenario, for a set of traffic flows arriving at the IoT terminal and requesting different network services, an ILP model Ps including loop prevention constraints is well formulated. An approximation algorithm AFGC with competitive ratio O(1+(|R|-1)α), α (0, 1) is then designed, which thoroughly search all key instances to run comprehensive loop break. In dynamic scenario for a flow request on the fly reaching an IoT terminal, an ILP model Pd and another approximation algorithm DFGU with a competitive ratio O(1 + |SCr||Fr| ) are developed, giving priority to reuse of existing topology to generate an augmented VNF-FG of minimum size. Simulation results indicate AFGC and DFGU outperform state-of-art work, and the gap between each of the two algorithms and their respective ILP models is marginal.
- Open Access
- 7 October 2023
- Published by MDPI AG in Journal of Marine Science and Engineering
- Vol. 11 (10) , 1934
Dynamic Modeling and Robust Trajectory Tracking Control of a Hybrid Propulsion-Based Small Underwater Robot
This paper proposes a hybrid propulsion-based small underwater robot for robust trajectory tracking control in a harsh and complex underwater environment. The robot is equipped with a Coanda-effect jet thruster and a pair of propeller-based reconfigurable magnetic-coupling thrusters, allowing it to traverse safely in confined or cluttered spaces as well as cruise efficiently in the open water. To investigate the robot dynamic modeling, we first formulated its simplified mathematical model and estimated the hydrodynamic coefficients by performing the planar motion mechanism using CFD (computational fluid dynamics) simulation. Then, a double-loop trajectory tracking control architecture was designed considering the model uncertainties and environmental disturbances. Based on Lyapunov theory, the outer-loop kinematic control produces the virtual velocity command, while the inner-loop dynamic control adopts the full-state feedback L1-adaptive control to match the command. The asymptotic convergence of the tracking errors and the stability of the whole closed-loop system are guaranteed. Finally, comparative simulations in the presence of unknown disturbances and the variation of model parameters were carried out to verify the robustness of our proposed trajectory tracking control, which is also suitable for the separated son robots.
- 7 October 2023
- Published by MDPI AG in Journal of Marine Science and Engineering
- Vol. 11 (10) , 1934
Open AccessDynamic Modeling and Robust Trajectory Tracking Control of a Hybrid Propulsion-Based Small Underwater Robot
This paper proposes a hybrid propulsion-based small underwater robot for robust trajectory tracking control in a harsh and complex underwater environment. The robot is equipped with a Coanda-effect jet thruster and a pair of propeller-based reconfigurable magnetic-coupling thrusters, allowing it to traverse safely in confined or cluttered spaces as well as cruise efficiently in the open water. To investigate the robot dynamic modeling, we first formulated its simplified mathematical model and estimated the hydrodynamic coefficients by performing the planar motion mechanism using CFD (computational fluid dynamics) simulation. Then, a double-loop trajectory tracking control architecture was designed considering the model uncertainties and environmental disturbances. Based on Lyapunov theory, the outer-loop kinematic control produces the virtual velocity command, while the inner-loop dynamic control adopts the full-state feedback L1-adaptive control to match the command. The asymptotic convergence of the tracking errors and the stability of the whole closed-loop system are guaranteed. Finally, comparative simulations in the presence of unknown disturbances and the variation of model parameters were carried out to verify the robustness of our proposed trajectory tracking control, which is also suitable for the separated son robots.
- 10 October 2022
- Published by Institute of Electrical and Electronics Engineers (IEEE) in IEEE Sensors Journal
- Vol. 22 (23) , 23356-23366
The Deep Reinforcement Learning-Based VR Training System With Haptic Guidance for Catheterization Skill Transfer
With growing demands in endovascular intervention, priority is expected to be given to the catheter manipulation skill training. For novices, it’s inevitable to face the dilemma that progress is achieved slowly when they practice repeatedly. Thus, exploring underlying behavior patterns from experts’ demonstrations and transferring manipulation skills to novice are of significance when designing catheterization training systems. To address such issue, a deep reinforcement learning based VR training system has been presented, which aims to exploit expert’s experience-related skills to correct novices’ actions via the 2-DOF haptic guidance. The training system mainly consists of three parts: a force sensors integrated catheter manipulator, a VR simulator and a learning based model. As experience-related skills are encoded as the correlation between operator-tool and tool-tissue interactions, the deep deterministic policy gradient (DDPG), one of deep reinforcement learning algorithms is adopted so that expert’s skills can be extracted as the optimal policy model by learning from the catheterization demonstrations. Based on such learning model, the catheter manipulator can provide the intuitive haptic guidance to online correct novices’ catheterization maneuvers both in translational and rotational directions. To evaluate the skill improvement, a five-day training session has been carried out and results reveal that prominent progress has been made with respect to safety operation and task proficiency. Moreover, substantial evidences support the finding that the newly developed training system is qualified to transfer experience-related skills from expert to novice because trainees are capable of making wise decisions independently without any haptic cues after the training session
- 10 October 2022
- Published by Institute of Electrical and Electronics Engineers (IEEE) in IEEE Sensors Journal
- Vol. 22 (23) , 23356-23366
The Deep Reinforcement Learning-Based VR Training System With Haptic Guidance for Catheterization Skill Transfer
With growing demands in endovascular intervention, priority is expected to be given to the catheter manipulation skill training. For novices, it’s inevitable to face the dilemma that progress is achieved slowly when they practice repeatedly. Thus, exploring underlying behavior patterns from experts’ demonstrations and transferring manipulation skills to novice are of significance when designing catheterization training systems. To address such issue, a deep reinforcement learning based VR training system has been presented, which aims to exploit expert’s experience-related skills to correct novices’ actions via the 2-DOF haptic guidance. The training system mainly consists of three parts: a force sensors integrated catheter manipulator, a VR simulator and a learning based model. As experience-related skills are encoded as the correlation between operator-tool and tool-tissue interactions, the deep deterministic policy gradient (DDPG), one of deep reinforcement learning algorithms is adopted so that expert’s skills can be extracted as the optimal policy model by learning from the catheterization demonstrations. Based on such learning model, the catheter manipulator can provide the intuitive haptic guidance to online correct novices’ catheterization maneuvers both in translational and rotational directions. To evaluate the skill improvement, a five-day training session has been carried out and results reveal that prominent progress has been made with respect to safety operation and task proficiency. Moreover, substantial evidences support the finding that the newly developed training system is qualified to transfer experience-related skills from expert to novice because trainees are capable of making wise decisions independently without any haptic cues after the training session
- 6 July 2022
- Published by Institute of Electrical and Electronics Engineers (IEEE) in IEEE Transactions on Network and Service Management
- Vol. 19 (3) , 3371-3386
On Efficient Delay-Aware Multisource Multicasting in NFV-Enabled Softwarized Networks
To comply with security and performance policies, multicast communication requires inline service that chains an ordered sequence of virtualized network functions (VNFs) with an emerging paradigm of network function virtualization (NFV). As many-to-many multicast pattern is widely used especially in current MEC, multimedia and big data industries, in this paper we focus on multi-source NFV-enabled multicasting in software-defined networks (SDN). By jointly investigating VNF placement and multicast tree routing strategy, it is envisioned to construct service function tree with minimum network resource expenditure while yielding lower delay. We explore this problem in static and online scenarios individually. In static situation with a bunch of multicast requests to be served, a column generation based computing model is first formulated, based on which an approximation algorithm MDNM that leverages logic of Viterbi and puts emphasis on resource optimization and segmental delay satisfaction is developed. In online situation with a single incoming request, another ILP model to enhance VNF instance reuse is established. Then, SVM algorithm giving importance to source determination and VNF instance consolidation is designed to efficiently construct a service function tree with minimum resource usage and delay. Simulation results show that the gap between each of the two proposed algorithms and their respective ILP models is marginal, and both MDNM and SVM outperform the state-of-art work.
- 6 July 2022
- Published by Institute of Electrical and Electronics Engineers (IEEE) in IEEE Transactions on Network and Service Management
- Vol. 19 (3) , 3371-3386
On Efficient Delay-Aware Multisource Multicasting in NFV-Enabled Softwarized Networks
To comply with security and performance policies, multicast communication requires inline service that chains an ordered sequence of virtualized network functions (VNFs) with an emerging paradigm of network function virtualization (NFV). As many-to-many multicast pattern is widely used especially in current MEC, multimedia and big data industries, in this paper we focus on multi-source NFV-enabled multicasting in software-defined networks (SDN). By jointly investigating VNF placement and multicast tree routing strategy, it is envisioned to construct service function tree with minimum network resource expenditure while yielding lower delay. We explore this problem in static and online scenarios individually. In static situation with a bunch of multicast requests to be served, a column generation based computing model is first formulated, based on which an approximation algorithm MDNM that leverages logic of Viterbi and puts emphasis on resource optimization and segmental delay satisfaction is developed. In online situation with a single incoming request, another ILP model to enhance VNF instance reuse is established. Then, SVM algorithm giving importance to source determination and VNF instance consolidation is designed to efficiently construct a service function tree with minimum resource usage and delay. Simulation results show that the gap between each of the two proposed algorithms and their respective ILP models is marginal, and both MDNM and SVM outperform the state-of-art work.
- Open Access
- 20 January 2022
- Published by ALife Robotics Corporation Ltd. in Proceedings of International Conference on Artificial Life and Robotics
- Vol. 27, 416-419
Consensus Control of Linear Discrete-time Multi-agent Systems with Limited Communication Data Rate
- 20 January 2022
- Published by ALife Robotics Corporation Ltd. in Proceedings of International Conference on Artificial Life and Robotics
- Vol. 27, 416-419
Open AccessConsensus Control of Linear Discrete-time Multi-agent Systems with Limited Communication Data Rate
- Open Access
- 20 January 2022
- Published by ALife Robotics Corporation Ltd. in Proceedings of International Conference on Artificial Life and Robotics
- Vol. 27, 420-423
Formation control for rectangular agents with communication maintenance and collision avoidance
- 20 January 2022
- Published by ALife Robotics Corporation Ltd. in Proceedings of International Conference on Artificial Life and Robotics
- Vol. 27, 420-423
Open AccessFormation control for rectangular agents with communication maintenance and collision avoidance