Freeway Ramp Metering Using Fuzzy Logic and Genetic Proportional plus Integral Control

Xinrong Liang, Qi Lu, Peiqun Lin, Jianmin Xu
2016 International Journal of Control and Automation  
Past research has produced various ramp metering approaches. Even so, the results of existing ramp control techniques are unsatisfactory due to the highly nonlinear property of the ramp control system. For improving the ramp metering performance and raising the ramp control accuracy, we propose a mixed approach of fuzzy logic control and proportional plus integral (PI) control for freeway ramp metering. We also employ a genetic algorithm (GA) to optimize the PI parameters in this paper.
more » ... ore, we apply a nonlinear control technique with a feedback loop to enhance system performance. Firstly, a first-order traffic flow model called hydrodynamic model is established, and the model characteristics are analyzed. Secondly, the control objective of the ramp metering system is defined by the mainline traffic density. Thirdly, a mixed controller of fuzzy logic and genetic PI is designed based upon this hydrodynamic model and in combination with a nonlinear control technique. The membership functions of the fuzzy logic control are triangular or Gaussian curves, and the numbers of fuzzy control rules are nine. Fourthly, optimization procedures of a GA for finding the best PI control parameters are given with details. Finally, the authenticity and effectiveness of the mixed controller are verified with Matlab software R2010a and also by VISSIM microscopic traffic simulation. Simulation results prove that this mixed control approach has better tracking performance and smaller density errors compared with the method of artificial neural networks. This mixed control approach, as well as the nonlinear control technique, provides a new idea for freeway ramp metering. metering algorithms into two categories. One is the fixed-time metering, and the other is the traffic responsive metering. Metering rates in traffic responsive metering are determined based on real-time traffic status. In contrast, the fixed-time metering cannot respond to current traffic conditions. Consequently, traffic responsive metering is superior to the fixed-time metering in improving the freeway efficiency and in eliminating jams. Nowadays, intelligent control techniques are widely applied in the traffic responsive metering. As an intelligent control technique, artificial neural networks are mostly used in freeway traffic control. However, neural networks have some shortcomings in applying to freeway ramp metering. For instance, the training procedures, as well as the training data, have an influence on the network performance. Besides, the system parameters, such as the number of learning epochs and the number of neurons in hidden layers, also exert a tremendous influence on the network performance. Ramp metering system is an extremely nonlinear control system. For this reason, we use a nonlinear feedback control technique to design the traffic responsive metering algorithm so that we can improve the system performance. For raising the ramp control accuracy, we present a mixed approach of fuzzy logic control and PI control for freeway ramp metering. Also, we employ a GA to optimize the PI parameters. Because fuzzy logic control can easily construct nonlinear controllers, it is thought to be a promising method in many engineering fields. Fuzzy logic control has its rationality since its control strategies are similar to the operational process of experienced operators. Some rules of conditions and actions can state this working process. The conditions use linguistic terms like small or big, and the action representation uses linguistic terms like decrease marginally. Similarly, the fuzzy logic control uses if-then rules. However, fuzzy logic control has an equivalent effect of proportional plus differential (PD) control, and its steady-state deviation generated by disturbance signals may be not zero. Hence, we add an integral control to remove this offset. This mixed control approach combined with a nonlinear feedback control technique has several advantages. First, it can directly deal with a nonlinear problem and needs not to linearize the first-order traffic flow model. Second, GA can online adjust PI control parameters so that the system can adapt to constantly changing environments. Third, it has excellent tracking performance and tiny density errors. Equation (6) can completely describe the evolution process of the traffic stream in freeway mainlines.
doi:10.14257/ijca.2016.9.11.06 fatcat:mqhbactzbnfnhmjtdd66qwzoyy