Xiaoting Dong

Dr. Xiaoting Dong

orcid

Research Focus & Metrics

Academic Career Length

8Years (2019-Present)

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

Representative Publications

View All Publications
17 March 2025 in Complex & Intelligent Systems
Verified
Journal Article
Open access
PDF file
XDXiaoting Dong
Xiaoting Dong
GWguangxi wan
guangxi wan
PZPeng Zeng
Peng Zeng
Automated guided vehicles (AGVs) are widely used for transportation in flexible job shop (FJS) systems, and their transportation task scheduling has the same substantial impact on production efficiency as machine scheduling does. However, traditional FJS scheduling methods often prioritize job sequencing and machine selection while ignoring the impact of AGV transportation, resulting in suboptimal scheduling solutions and even difficulties in implementation. To address this issue, this paper formulates a cooperative scheduling model by introducing the AGV scheduling problem into the classical FJS scheduling problem, abbreviated as the FJS-AGV problem, with the objective of minimizing the makespan. With respect to the FJS-AGV problem, a heuristic-assisted deep Q-network (HA-DQN) algorithm is proposed, which leverages heuristic rules to enable the decision agent to perform multiple actions at each decision point, which includes determining the responses to the following questions: Which operation should be processed next? On which machine? By which AGV? This decision mechanism enables the agent to make more informed decisions, leading to improved performance and resource allocation in the FJS-AGV system. The practicability of the proposed FJS-AGV model and the efficiency of the HA-DQN algorithm in solving the FJS-AGV problem are verified through various international benchmarks. Specifically, when solving instances in a large benchmark, the HA-DQN algorithm yields a significant 12.63% reduction in makespan compared with that when traditional heuristics are employed.

27 December 2023 in Biomimetics
Verified
Journal Article
Open access
PDF file
XDXiaoting Dong
Xiaoting Dong
GWguangxi wan
guangxi wan
PZPeng Zeng
Peng Zeng
CSChunhe Song
Chunhe Song
SCShijie Cui
Shijie Cui
The robot task sequencing problem and trajectory planning problem are two important issues in the robotic optimization domain and are solved sequentially in two separate levels in traditional studies. This paradigm disregards the potential synergistic impact between the two problems, resulting in a local optimum solution. To address this problem, this paper formulates a co-optimization model that integrates the task sequencing problem and trajectory planning problem into a holistic problem, abbreviated as the robot TSTP problem. To solve the TSTP problem, we model the optimization process as a Markov decision process and propose a deep reinforcement learning (DRL)-based method to facilitate problem solving. To validate the proposed approach, multiple test cases are used to verify the feasibility of the TSTP model and the solving capability of the DRL method. The real-world experimental results demonstrate that the DRL method can achieve a 30.54% energy savings compared to the traditional evolution algorithm, and the computational time required by the proposed DRL method is much shorter than those of the evolutionary algorithms. In addition, when adopting the TSTP model, a 18.22% energy reduction can be achieved compared to using the sequential optimization model.

16 November 2023 in IEEE Robotics and Automation Letters
Verified
Journal Article
Closed access
QDQingwei Dong
Qingwei Dong
PZPeng Zeng
Peng Zeng
GWguangxi wan
guangxi wan
YHYunpeng He
Yunpeng He
XDXiaoting Dong
Xiaoting Dong
Deepreinforcement learning algorithms offer a promising method for industrial robots to tackle unstructured and complex scenarios that are difficult to model. However, due to constraints related to equipment lifespan and safety requirements, acquiring a number of samples directly from the physical environment is often infeasible. With the development of increasingly realistic simulators, it has become feasible for industrial robots to acquire complex motion skills within simulated environments. Nonetheless, the “reality gap” frequently results in performance degradation when transferring policies trained in simulators to physical systems. In this letter, we treat the reality gap between a physical environment (target domain) and a simulated environment (source domain) as a Gaussian perturbation and utilize Kalman filtering to reduce the discrepancy between source and target domain data. We refine the source domain controller using target domain data to enhance the controller's adaptability to the target domain. The efficacy of the proposed method is demonstrated in reaching tasks and peg-in-hole tasks conducted on PR2 and UR5 robotic platforms.

29 October 2024 in IEEE Robotics and Automation Letters
Verified
Journal Article
Closed access
QDQingwei Dong
Qingwei Dong
PZPeng Zeng
Peng Zeng
YHYunpeng He
Yunpeng He
GWguangxi wan
guangxi wan
XDXiaoting Dong
Xiaoting Dong
Complex operational scenarios increasingly demand that industrial robots sequentially resolve multiple interrelated problems to accomplish complex operational tasks, necessitating robots to have the capacity for not only learning through interaction with the environment but also for continual learning. Current deep reinforcement learning methods have demonstrated substantial prowess in enabling robots to learn individual simple operational skills. However, catastrophic forgetting regarding the continual learning of various distinct tasks under a unified control policy remains a challenge. The lengthy sequential decision-making trajectory in reinforcement learning scenarios results in a massive state-action search space for the agent. Moreover, low-value state-action samples exacerbate the difficulty of continuous learning in reinforcement learning problems. In this letter, we propose a Continual Reinforcement Learning (CRL) method that accommodates the incremental multiskill learning demands of robots. We transform the tightly coupled structure in Guided Policy Search (GPS) algorithms, which closely intertwine local and global policies, into a loosely coupled structure. This revised structure updates the global policy only after the local policy for a specific task has converged, enabling online learning. In incrementally learning new tasks, the global policy is updated using hard parameter sharing and Memory Aware Synapses (MAS), creating task-specific layers while penalizing significant parameter changes in shared layers linked to prior tasks. This method reduces overfitting and mitigates catastrophic forgetting in robotic CRL. We validate our method on PR2, UR5 and Sawyer robots in simulators as well as on a real UR5 robot.

24 June 2024 in Robotics and Computer-Integrated Manufacturing
Verified
Journal Article
Closed access
XDXiaoting Dong
Xiaoting Dong
GWGuangxi Wan
Guangxi Wan
PZPeng Zeng
Peng Zeng
CSChunhe Song
Chunhe Song
SCShijie Cui
Shijie Cui
YLYiyang Liu
Yiyang Liu

1 January 2019
Verified
Conference Paper
Open access
PDF file
XDXiaoting Dong
Xiaoting Dong
Based on the strategic planning of Shanghai Metro Smart Station and the main problems existing in the Automatic Fare Collection System of rail transit, the application requirements of speech recognition technology in the Automatic Fare Collection System of rail transit are put forward. On the basis of construction background, research and development of key technology points, software and hardware architecture of the scheme and pilot analysis, a voice ticket purchase scheme is proposed. On the basis of the original ticket purchase process of Ticket Vending Machine (TVM), voice recognition technology is used to increase the functions of wake-up-free voice input, Chinese phonetic alphabet input and fuzzy location inquiry, so as to reduce the queue pressure of stations with large passenger flow, improve the efficiency of passenger ticket purchase, and the degree of equipment modernization.