Coordination Concepts for Ramp Metering Control in a Freeway Network

The goal of this paper is to describe a generic approach to the problem of optimal coordinated ramp metering control in large-scale motorway networks. In this approach, the traffic flow process is macroscopically modeled by use of a second-order macroscopic traffic flow model. The overall problem of coordinated ramp metering is formulated as a constrained discrete-time nonlinear optimal control problem, and a feasible-direction nonlinear optimization algorithm is employed for its numerical solution. The control strategy's efficiency is demonstrated through its application to the 32-km Amsterdam ring road. A number of adequately chosen scenarios along with a thorough analysis, interpretation, and suitable visualization of the obtained results provides a basis for the better understanding of some complex interrelationships of partially conflicting performance criteria. More precisely, the strategy's efficiency and equity properties as well as their tradeoff are studied and their partially competitive behavior is discussed. The results of the presented approach are very promising and demonstrate the efficiency of the optimal control methodology for motorway traffic control problems.

Ramp metering has emerged as a freeway traffic management control strategy that facilitates efficient freeway operations. It has been recognized as an effective freeway management strategy that can significantly improve freeway capacity utilisation as well as increase overall network performance. However, despite these potential benefits, a number of limitations and challenges in developing ramp metering systems have been revealed from the literature. These include equity between road users, performance during over-saturated condition, real-time responsiveness and the challenges of a decentralised control approach. This thesis aims to develop and evaluate the proposed multi-agent based ramp metering algorithm and investigate some important aspects arising from the implementation of ramp metering systems. Agent technologies were introduced to model an individual ramp controller that would react dynamically under real-time traffic information. The agent concept of cooperation and collaboration was applied to develop a decentralised ramp metering system to balance system-wide efficiency and equity between road users. The proposed multi-agent ramp metering control was modelled as a hierarchical multi-agent control system comprising three operational control levels - local, group, and region. Each controller was modelled as an individual agent. At each control level, different goals and principles in metering rate calculation were applied. The developed multi-agent ramp metering algorithm (AGENT) was comparatively evaluated with other control algorithms including no control, time-of-day plan, ALINEA, FLOW and Stratified Zone under recurring and non-recurring traffic congestion. The developed Pacific Motorway traffic simulation model was applied as the study test-bed for evaluation of the algorithm performances. Under recurring traffic congestion, AGENT was found to perform very well under both normal (100 percent) and heavy (120 percent) traffic demand. Under normal traffic demand, it was found that AGENT was obviously superior to FLOW and Stratified Zone, but slightly worse than ALINEA (1.6 percent less on average flow) in terms of overall network performance. However, for on-ramp performance, AGENT was found to increase on-ramp flow by 20 percent compared to ALINEA. It was also found that AGENT generated the lowest average travel time, lowest travel delay and least on-ramp waiting time, all of which are the most important indices in term of network equity. The benefit of AGENT was found when traffic demand increased (heavy demand). It was found from the experiment that AGENT was considerably superior to ALINEA (1.7 percent higher on average flow) and other algorithms. In term of equity, AGENT was found to have the ability to balance network efficiency and equality by producing high network efficiency and excellent equity to the road users. According to the experiments under non-recurring traffic congestion regardless of severity and duration of incident, AGENT was found to have superior performance compared to other control strategies, including ALINEA. In addition, it can significantly recover the impact from incident faster than other algorithms. Its network-wide benefits over those of ALINEA, FLOW and Stratified Zone are up to 1.3, 11 and 16.1 percent on the average flow, up to 2.7, 38.7 and 67.9 percent on average speed and up to 7.4, 20.9 and 17.1 percent on average delay, respectively. For the on-ramp performance, the benefits of AGENT over ALINEA and FLOW can be found from the improvement on average on-ramp flow by 20.3, 2.4 percent and the reduction on average on-ramp delay by 1.8, 33.9 percent, respectively. Compared with Stratified Zone, AGENT was found to increase average on-ramp speed by up to 16 percent. These combined findings clearly demonstrate the potential of a multi-agent ramp metering algorithm in enhancing freeway performance and reducing congestion. This thesis has successfully achieved its stated objectives by demonstrating the feasibility of applying a decentralised approach to a ramp metering control system using a multi-agent concept and evaluating its performance against other well-known ramp metering algorithms. The study also successfully achieved its secondary objectives which included advancing the state of knowledge in the calibration and validation of traffic simulation model for freeway network; enhancing the procedures for ramp metering rate determination; formulating methodologies and frameworks for determining and calibrating parameters for ramp metering algorithm, and formulating general methodologies and frameworks that can be used to evaluate the performance of Intelligent Transport Systems applications.

The problem of traffic congestion in modern day motorways calls for the design and implementation of efficient control strategies. It is argued in this paper that in order to have efficient, generic, and systematic solutions to a wide range of traffic control problems, macroscopic motorway traffic flow models in state-space form, that are relevant for the control problem and computationally non-intensive, are most appropriate. Such models allow the exploitation of available powerful, systematic, and theoretically supported automatic control concepts. Based on these concepts an Extended Kalman Filter for traffic state estimation, a multivariable LQI controller for coordinated ramp metering on a motorway stretch, and an integrated optimal control strategy for motorway networks are shortly presented. The criteria of a model's relevance for a given traffic control problem and its computational requirements are subsequently examined. Finally, the application of an advanced coordinated ramp metering control strategy, based on the optimal control approach, to the ring-road of Amsterdam, The Netherlands, is provided as an illustrative example.

Network Impact of Dynamic Traffic Management Measures

Traffic state observers derived from the cell transmission model (CTM) are vulnerable to incorrect information and time variation of traffic flow parameters, in particular the critical density, known as the issue of mismodelling, which can cause erroneous mode switching of CTM-based observers to undermine estimation and control. This issue cannot be completely overcome, even if traffic flow parameters are augmented into state vector. We propose coupling a supervisor to the standard CTM-EKF observer to completely resolve this issue. Simulation shows that the proposed supervised observer can switch modes in accordance with actual situations and generate quality estimates of both traffic state and parameters. The supervised observer is then integrated with ramp metering control. Simulation shows that, in an environment of time-varying traffic flow parameters, the supervised observer-based ramp metering control system considerably outperforms an ordinary observer-based ramp metering control system, which only updates traffic state in real time.

SUMMARYIn this paper, a probe‐based adaptive ramp metering approach has been developed and compared with a detector‐based and pretimed ramp metering approaches using PARAMICS microsimulation model made by Quadstone Limited. Both examined probe‐based and detector‐based ramp metering approaches adopt the same problem formulation; however, they differ by the data source that provides estimates of their input parameters. The probe‐based approach takes as main input space mean speed extracted from vehicle probes moving constantly on the entire freeway. The detector‐based approach takes the density readings directly from point detectors placed upstream and downstream of each on‐ramp. The developed approach formulates the ramp metering rates to simultaneously reduce freeway travel time and decrease delays on ramps. A dynamic weighting scheme is adopted in this formulation to give different weighting schemes to the various freeway sections and the on‐ramps. These weights are able to adjust dynamically based on prevailing and forecasted traffic conditions to give priority in each step to the sections of the freeways or/and on‐ramps experiencing critical traffic conditions. The comparative analysis of the three ramp metering approaches was conducted on an 8‐km freeway section on Highway 2 in Calgary, Alberta, Canada. The results of the analysis indicate that the probe‐based algorithm consistently outperformed the two other approaches in terms of all performance measures. The probe‐based approach is shown to exceed the performance of the detector‐based approach with a market share of vehicle probes as low as 3%. The results of the analysis are promising and show the effectiveness of developing ramp metering control algorithms taking the main input parameters from vehicle probes on the freeway. Copyright © 2012 John Wiley & Sons, Ltd.

Multicriteria ramp metering control considering environmental and traffic aspects

Ramp metering is one of the successful active traffic control strategies to control traffic flow at entry points to freeways. This study evaluates the effectiveness of fixed-time ramp metering control strategy on the day-to-day operation of traffic over two segments (easternmost and westernmost) of the I-12 corridor in Baton Rouge, Louisiana. Detector speeds and volumes were collected over 11 months and used to generate three performance measures, speed, travel time, and level of service to compare traffic conditions before and after the deployment of ramp meters. Comparative analysis, comprising statistical analysis, analysis of travel time savings, and level of service were then undertaken for traffic conditions before and after ramp meters installation. Overall, the results show some improvements in traffic conditions in the eastbound direction of the westernmost segment; however, the conditions slightly deteriorated on the westbound of the same segment. For the easternmost segment of I-12, no improvement was detected.

A physics-informed reinforcement learning-based strategy for local and coordinated ramp metering

Considered in this paper is a novel model-based, coordinated ramp metering strategy. It aims at maximizing the discharge flow in motorway networks by minimizing the divergence of the traffic density from its critical value caused by unknown demand flow. The suggested synthesis algorithm casts the traffic control objective into the form of an induced $\mathcal {L}_{2}$ -norm minimization problem. Hence, we aim at rejecting the effect of disturbance on the overall network performance output while the ramp input flow is subjected to constraints. With such a problem formulation, it is not required to know the disturbance input in order to find the proper control input. Without any central decision unit (traffic control center), ramp meters coordinate by sharing their local variables with solely their neighbor units (upstream and downstream) to achieve the global performance goal. Under some network symmetry conditions, a compositionally inexpensive distributed flow control method is suggested to address scalability issues. The method is implemented in simulation environment and compared with other control algorithms in two comprehensive case studies.

Taking Beijing urban expressway as research object, the optimal traffic control strategies on model predictive control for ramp metering in urban expressway is considered in this paper. A generic problem formulation is presented in the format of a discrete-time optimal control problem whose numerical solution is achieved by use of the minimal principle on discrete-time non-linear system. An illustrative example with two pairs of entrance ramp and exit ramp is validated and the results demonstrate the efficiency of the approach proposed.

ABSTRACTThis paper presents the design and evaluation process of a self-learning system for local ramp metering control. This system is developed on the basis of reinforcement learning (RL) and can deal with the problem of on-ramp queue management through a continuous learning process. A general framework of the system design including the definition of RL elements and an algorithm that can accomplish the learning process is proposed. Simulation tests are carried out to evaluate the performance of the new system. In terms of the total time spent by road users, the new system can achieve a 30% reduction from the situation of no control, a result which is competitive with the widely accepted algorithm ALINEA. Meanwhile, simulation results show that the new system can keep on-ramp queues strictly under a series of pre-specified constraints, which proves its capability of managing on-ramp queues.

The development of on ramp metering control strategies for a freeway can be aided by the well-developed theory of linear optimal control. The recently developed variable model appears to be best suited for this purpose. In this model the freeway is partitioned into lengths of about 1/2 mile and traffic is described in terms of the aggregate variables section speed, section density, and volume. This model and linear optimal control theory are used to developed an on ramp metering strategy which is designed to maintain freeway conditions near nominal conditions in the presence of disturbances. An illustrative design for a segment of the San Diego freeway in Los Angeles is provided. The companion problem of processing freeway sensor data to provide current estimates of the aggregate variables can be approached through standard techniques in estimation theory. Methods of attacking this portion of the freeway surveillance problem are presented.

This paper proposes a new ramp metering control algorithm, Feed-Foward ALINEA (FF-ALINEA), for bottlenecks located both nearby on an on-ramp and further away from it (i.e., more than just a few hundred meters). The formulation of the controller is based on a feed-forward modification of the well-known control algorithm for ramp metering, ALINEA. The feed-forward structure allows anticipating the future evolution of the bottleneck density in order to avoid or reduce traffic breakdowns. The proposed controller is tested, using the macroscopic traffic flow model METANET, for nine scenarios, and the results are compared with the ones obtained with ALINEA, PI-ALINEA, and with the optimal solution. The simulations show that the FF-ALINEA is able to approach the optimal behavior, thereby outperforming ALINEA and PI-ALINEA. Moreover, results indicate that the FF-ALINEA is quite robust in cases where different demands are considered, there are a limited number of available detectors, or there are errors in the estimation of the capacity and/or the critical density of the bottleneck.

In order to solve the traffic congestion control problem on expressway, the influencing areas in the upstream of the congestion section and a kind of heuristic coordinated ramp metering methods are focused on. According to the mechanism that traffic flow moves to the upstream of the section in the congestion duration time, queue length is considered as the significant factor in the determination of congestion influencing areas. The formula is deduced to calculate the number of ramps in the influencing areas. Due to the low adjusting speed in the single point level and wide vibration amplitude in the coordination level in the traditional methods, the density is selected as the control parameter instead of the traffic flow volume in the Bottleneck algorithm. Meanwhile, the improved on ramp coordination metering method and the adjustment rate formula are formed based on Bottleneck algorithm eliminating the limitations of traditional methods. Taken the road section from Gongyi to Luoyang on the Lianhuo expressway as the instance, coordinated ramp metering simulation analysis is conducted by using the MATLAB software. The rationality and effectiveness of the formula of calculating the number of ramps in the congestion influencing areas and the improved algorithm are demonstrated by means of the numerical simulation.