Li Fang

Dr. Li Fang


Research Focus & Metrics

Academic Career Length

11Years (2016-Present)

(Calculated once Scilit publication data is synced: current year − year of first publication + 1)

Representative Publications

View All Publications
19 April 2018 in IEEE Access
Verified
Journal Article
Open access
XCXinye Chen
Xinye Chen
PZPing Zhang
Ping Zhang
GDGuanglong Du
Guanglong Du
LFLi Fang
Li Fang
This paper considers the problem of having a team of mobile robots to visit a set of target locations. This problem is known as multi-robot patrolling problems. In this paper, the problem is formulated as a multiple traveling salesman problem (MTSP) with single depot or multiple depot, which is an nondeterministic polynomial-hard problem. Unlike most previous research works, in real-world applications, the requirement of optimizing the maximum traveled distance and the total traveled distance simultaneously widely exists. In this paper, a bi-objective ant colony optimization (ACO) based memetic algorithm is proposed to solve the problem. In the algorithm, a simple multi-ACO is integrated with a sequential variable neighborhood descent. A powerful local optimization method for bi-objective MTSP is proposed to improve the candidate solutions. In addition, we adopt the technique for order preference by similarity to an ideal solution method to select a reasonable solution from the optimal Pareto. Through computational experiments, we demonstrated the benefits of our algorithm as compared with four other existing algorithms. Computational results show that proposed algorithm is promising and effective for the bi-objective MTSPs.

7 November 2019 in IEEE Access
Verified
Journal Article
Open access
LZLanxiang Zheng
Lanxiang Zheng
PZPing Zhang
Ping Zhang
JTJia Tan
Jia Tan
LFLi Fang
Li Fang

12 April 2024 in Sustainability
Verified
Journal Article
Open access
PDF file
YPYu Pu
Yu Pu
LFLi Fang
Li Fang
Shahin RahimifardShahin Rahimifard
Shahin Rahimifard
In response to the challenges of dynamic adaptability, real-time interactivity, and dynamic optimization posed by the application of existing deep reinforcement learning algorithms in solving complex scheduling problems, this study proposes a novel approach using graph neural networks and deep reinforcement learning to complete the task of job shop scheduling. A distributed multi-agent scheduling architecture (DMASA) is constructed to maximize global rewards, modeling the intelligent manufacturing job shop scheduling problem as a sequential decision problem represented by graphs and using a Graph Embedding–Heterogeneous Graph Neural Network (GE-HetGNN) to encode state nodes and map them to the optimal scheduling strategy, including machine matching and process selection strategies. Finally, an actor–critic architecture-based multi-agent proximal policy optimization algorithm is employed to train the network and optimize the decision-making process. Experimental results demonstrate that the proposed framework exhibits generalizability, outperforms commonly used scheduling rules and RL-based scheduling methods on benchmarks, shows better stability than single-agent scheduling architectures, and breaks through the instance-size constraint, making it suitable for large-scale problems. We verified the feasibility of our proposed method in a specific experimental environment. The experimental results demonstrate that our research can achieve formal modeling and mapping with specific physical processing workshops, which aligns more closely with real-world green scheduling issues and makes it easier for subsequent researchers to integrate algorithms with actual environments.

6 November 2018 in Frontiers in Robotics and AI
Verified
Journal Article
Open access
GDGuanglong Du
Guanglong Du
SLShuaiying Long
Shuaiying Long
LFLi Fang
Li Fang
XHXin Huang
Xin Huang
With the development of Industry 4.0, the cooperation between robots and people is increasing. Therefore, man—machine security is the first problem that must be solved. In this paper, we proposed a novel methodology of active collision avoidance to safeguard the human who enters the robot's workspace. In the conventional approaches of obstacle avoidance, it is not easy for robots and humans to work safely in the common unstructured environment due to the lack of the intelligence. In this system, one Kinect is employed to monitor the workspace of the robot and detect anyone who enters the workspace of the robot. Once someone enters the working space, the human will be detected, and the skeleton of the human can be calculated in real time by the Kinect. The measurement errors increase over time, owing to the tracking error and the noise of the device. Therefore we use an Unscented Kalman Filter (UKF) to estimate the positions of the skeleton points. We employ an expert system to estimate the behavior of the human. Then let the robot avoid the human by taking different measures, such as stopping, bypassing the human or getting away. Finally, when the robot needs to execute bypassing the human in real time, to achieve this, we adopt a method called artificial potential field method to generate a new path for the robot. By using this active collision avoidance, the system can achieve the purpose that the robot is unable to touch on the human. This proposed system highlights the advantage that during the process, it can first detect the human, then analyze the motion of the human and finally safeguard the human. We experimentally tested the active collision avoidance system in real-world applications. The results of the test indicate that it can effectively ensure human security.

3 December 2024 in Applied Soft Computing
Verified
Journal Article
Closed access
WGWeisen Guo
Weisen Guo
LFLi Fang
Li Fang
PZPing Zhang
Ping Zhang
LLLong Luo
Long Luo

1 August 2023 in IEEE Transactions on Neural Networks and Learning Systems
Verified
Journal Article
Closed access
HZHongliang Zeng
Hongliang Zeng
PZPing Zhang
Ping Zhang
LFLi Fang
Li Fang
CLChubin Lin
Chubin Lin
JZJunkang Zhou
Junkang Zhou
With shaped reward functions, reinforcement learning (RL) has recently been successfully applied to several robot control tasks. However, designing a task-relevant and well-performing reward function takes time and effort. Still, if RL can train an agent to complete a task in a sparse reward environment, it is an effective way to address the difficulty of reward function design, but it is still a significant challenge. To address this issue, the pioneering hindsight experience replay (HER) method dramatically enhances the probability of acquiring skills in sparse reward environments by transforming unsuccessful experiences into helpful training samples. However, HER still requires a lengthy training period. In this article, we propose a new technique based on HER termed adaptive HER with goal-amended curiosity module (AHEGC) for further enhancing sample and exploration efficiency. Specifically, an adaptive adjustment strategy of hindsight experience (HE) sampling rate and reward weights is developed to enhance sample efficiency. Furthermore, we introduce a curiosity mechanism to encourage more efficient exploration of the environment and propose a goal-amended (GA) curiosity module as a solution to the problem of over-seeking novelty caused by the curiosity introduced. We conducted experiments on six demanding robot control tasks with binary rewards, including Fetch and Hand environments. The results show that the proposed method outperforms existing methods regarding learning ability and convergence speed.