Projects
TOC
Robotic Autonomous Trolley Collection
Distributed Consensus of Constrained MAS under Directed Graphs
Vision-Based Auto Parking
Six-Arm Robot Palletizing
Line Patrolling Robot
Stable Controller Design for Inverted Pendulum
Stable Controller Design for Magnetic Levitation
Centralized Multi-Robot Task Allocation
Robotic Autonomous Trolley Collection
(An autonomous robotic collecting system incorporating detection, safe navigation, manipulation, and transportation.)
- Advisors: Prof. Max Q.-H. Meng (Fellow, IEEE); Prof. Jiankun Wang
- The objective is to bring about a systematic solution to robotic autonomous trolley collection in complex and dynamic large-scale environments, like international airports.
- This integrated system renders itself a challenge involving object detection, estimation & prediction, localization, efficient & safe planning and control, and robust mechanical design.
- Please see this page or refer to our ICRA 2022 paper for details.
Distributed Consensus of Constrained MAS under Directed Graphs
(Simulation results powered by the Robotarium )
Advisor: Prof. Jie Mei (Google Scholar)
- We aim to provide a distributed leaderless consensus algorithm framework for a type of continuous-time linear multi-agent systems with time-varying asymmetric state constraints, uncertainties, and disturbances under time-varying directed topologies.
- Theoretical proof, numerical simulations and physical experiments are presented.
- Please see this page for details.
Vision-Based Auto Parking
- Advisor: Prof. Haoyao Chen
- Team: Jiahao Fang, Hao Luan, Weijie Wu.
- Identified a specific parking sign by adopting filtering, color segmentation, perspective transformation, Canny edge detection and rectangle envelope.
- Designed an online closed-loop controller to control angular and linear velocities of an autonomous car, by employing multiple control schemes and using image information of the detected parking sign.
- Integrated searching, detection, and motion control on ROS and successfully realized fully automated parking.
Six-Arm Robot Palletizing
- Advisor: Prof. Yunjiang Lou (Google Scholar)
- Team: Hao Luan, Fangcheng Zhu
- Designed robot manipulator control algorithms using forward and inverse kinematics and LFPB trajectory planning. Built position management system to add, store, modify and delete position information of objects.
- Achieved fast palletizing motions with high accuracy.
(Due to regulations of the laboratory, we were only allowed to operate the machine with 25% of its max speed, so this video is played with 4X fast forward.)
Line Patrolling Robot
- Advisor: Prof. Haoyao Chen
- Team: Hao Luan, Fangcheng Zhu.
- Designed and 3D-printed structural parts, built electric circuits connecting basic modules including H bridges and DC motors, and used Arduino microcontroller to realize feedback control of the speed of DC motors with encoders.
- Designed a hybrid PID control algorithm based on the infrared sensors to control the robot tracking the desired black line.
- Integrated the hardware system, tested the robot and the algorithm in a complex map, and finished the task in 30s.
(This video is only a test, not the final version of our robot.)
Stable Controller Design for Inverted Pendulum
- Advisor: Prof. Ai-Guo Wu
- Modeled and linearized the inverted pendulum system in transfer function model and state space model respectively.
- Designed a controller via second-, third-, and fourth-order state feedback respectively using poles placement techniques.
- Designed a controller via output feedback using the root locus method.
Stable Controller Design for Magnetic Levitation
- Advisor: Prof. Ai-Guo Wu
- Modeled the magnetic levitation system and obtained its linearized plant model in transfer function.
- Designed a feedback controller using the root locus method.
Centralized Multi-Robot Task Allocation
- Advisor: Prof. Hui Cheng @ SYSU
- Team: Hao Luan, Yanheng Wang, Zihao Zeng.
- Designed and implemented a centralized offline task-allocation algorithm for multi-robot systems based on the Ant Colony System.
- Compared with a conventional optimal DFS algorithm, simulation showed time-consuming ratios down to 1/2400 and relative errors below 10% in dense directed graphs consisting of up to 20 vertices.