2018 4th International Conference on Mechanical Engineering and Automation Science | |
On-Policy Learning for the Swing Process Control of a Cutter Suction Dredger | |
机械制造;原子能学 | |
Wei, Changyun^1 ; Chen, Xiujing^1 ; Ni, Fusheng^1 | |
College of Mechanical and Electrical Engineering, Hohai University, China^1 | |
关键词: Adaptive Control; Continuous State Space; Cutter suction dredger; Dynamic characteristics; Environmental conditions; Generalization capability; Linear neural network; Production efficiency; | |
Others : https://iopscience.iop.org/article/10.1088/1757-899X/470/1/012017/pdf DOI : 10.1088/1757-899X/470/1/012017 |
|
来源: IOP | |
【 摘 要 】
It is hard to describe the swing process of a cutter suction dredger with accurate mathematical models because the dynamic characteristics of the swing process are complex, and the relationship between the parameters is not clear. Currently, the swing process control depends entirely on the operators, and it is sometimes difficult to obtain a high production efficiency and construction accuracy. In this paper, an approach that combines SARSA-Lambda with a linear neural network is proposed to solve the intelligent control of the swing process. The dynamic model of the swing process is built using the generalization capability of the linear neural network that solves the output problem of the continuous state space. SARSA-Lambda is used to realize the adaptive control of the swing process by means of learning. The simulation results show that the proposed approach can quickly and effectively learn and achieve goals in uncertain environmental conditions and achieve better control consequents.
【 预 览 】
Files | Size | Format | View |
---|---|---|---|
On-Policy Learning for the Swing Process Control of a Cutter Suction Dredger | 723KB | download |