Ali Hajbabaie

Automated Speed Photo Enforcement Effects on Speeds in Work Zones

Automated speed enforcement in construction zones has the potential to increase compliance with the speed limit and improve safety. The effectiveness of speed photo enforcement (SPE) (by radar) in reducing speeds and increasing speed limit compliance in work zones was evaluated for the first time in the United States, at Illinois work zones. Details are presented on SPE implementation and its effectiveness at the point it was stationed and at a downstream location in a work zone. Speed data were collected at the location of SPE and at a location 1.5 mi downstream in the work zone to determine the point and spatial effects of SPE. Speeds were measured for free-flowing and platooned cars and heavy vehicles in shoulder and median lanes. Results showed that SPE is effective in reducing the average speed and increasing compliance with work zone speed limit. The SPE reduced speed in the median lane more than in the shoulder lane, as expected. In addition, the speed of free-flowing vehicles was reduced more than for platooned vehicles. The reduction of the mean speed varied from 3.2 to 7.3 mph. The percentage of vehicles exceeding the speed limit near SPE was reduced from about 40% to 8% for free-flowing cars and from 17% to 4% for free-flowing heavy vehicles. Near the SPE van, none of the cars exceeded the speed limit by more than 10 mph, and none of the heavy vehicles exceeded it by more than 5 mph. The data also showed a mixed spatial effect for SPE. At the downstream location, the speed reduction for cars was not significant, while it varied from 0.9 to 2.5 mph for heavy vehicles.

A Program for Simultaneous Network Signal Timing Optimization and Traffic Assignment

This study formulates a program for simultaneous traffic signal optimization and system optimal traffic assignment for urban transportation networks with added degree of realism. The formulation presents a new objective function, i.e., weighted trip maximization, and explicit constraints that are specifically designed to address oversaturated conditions. This formulation improves system-wise performance while locally prevents queue spillovers, de-facto reds, and gridlocks. A meta-heuristic algorithm is developed that incorporates microscopic traffic flow models and system optimal traffic assignment in genetic algorithms. This solution technique efficiently optimizes signal timing parameters, at the same time solves system optimal traffic assignment, and accounts for oversaturated conditions and different drivers behaviors. This study also proposes a framework to calculate an upper bound on the value of the objective function by solving the problem while several constraints (i.e., network loading and traffic assignment) are relaxed. An empirical case study for a portion of downtown Springfield, Illinois has been conducted under four demand patterns. Findings indicate that our solution approach can solve the problem effectively. Several managerial insights have also been drawn.

Downstream effects of speed photo-radar enforcement and other speed reduction treatments on work zones

The effects of automated speed photo–radar enforcement (SPE) and traditional speed reduction treatments (speed feedback trailer, presence of police vehicles with emergency lights on and off, and combinations of the speed feedback trailer and police presence) on speed were studied at a location 1.5 mi downstream of the actual treatment (spatial effects). Three data sets from two Interstate highway work zones were used. Field data consistently showed significant spatial (downstream) effects for SPE. The combination of speed feedback trailer and police vehicle with emergency lights off had downstream effects in some cases but to a lesser degree than SPE. Other treatments showed no significant downstream effects. For free-flowing traffic, SPE reduced the average downstream speed by 2 to 3.8 mph for cars and by 0.8 to 5.3 mph for trucks. Also, SPE reduced speeding cars by 7.1% to 23.4% (except for cars in median in Data Set 1), and speeding trucks by 4.2% to 48.3% (except for trucks in shoulder in Data Set 3). For the general traffic stream, SPE reduced the average downstream speed by 1.1 to 2.9 mph on cars and by 0.9 to 3.3 mph on trucks. When SPE was used, the percentage of speeding cars and trucks in the general traffic stream was reduced by 2.9% to 28.6%, and by 7.5% to 36.1%, respectively. SPE also reduced the percentage of cars in the general traffic stream exceeding the speed limit by more than 10 mph in virtually all cases, and eliminated such trucks in all but one case.

Speed Photo-Radar Enforcement and Its Effects on Speed in Work Zones

Automated speed photo–radar enforcement (SPE) in work zones was implemented for the first time in the United States in Illinois. This paper presents the results of the effectiveness of SPE on the basis of three data sets collected in two work zones. SPE was effective in reducing the average speed and increasing compliance with the work zone speed limit in all three data sets. In almost all cases in which SPE was implemented, the average speeds were significantly lower than the work zone speed limit. The average free-flowing speed of cars was reduced by 4.2 to 7.9 mph, and that of trucks by 3.4 to 6.9 mph. SPE reduced the percentage of cars and heavy vehicles exceeding the speed limit significantly. The percentages of free-flowing cars exceeding the speed limit were reduced from 39.8% to 8.3% in Data Set 1, from 30.4% to 4.2% in Data Set 2, and from 93.2% to 45.5% in Data Set 3. The percentages of free-flowing heavy vehicles exceeding the speed limit were reduced from 17.3% to 4.2% in Data Set 1; from 6.1% to 1.2% in Data Set 2; and from 69.2% to 13.9% in Data Set 3. Trucks did not exceed the speed limit by more than 10 mph in any of the data sets when SPE was implemented. In two data sets no cars exceeded the speed limit by more than 10 mph, while in the third data set only 2.5% did. Field data were also collected after the SPE van left the work zone to examine the halo (temporal) effects of SPE. SPE had a halo effect of 1.8∼2.7 mph on free-flowing trucks in one work zone but none in the other work zone. The halo effect of SPE on free-flowing cars was a limited 1.2 mph on the shoulder lane in only one data set.

Traffic Signal Timing Optimization

Choosing an appropriate objective function in optimizing traffic signals in urban transportation networks is not a simple and straightforward task because the choice likely will affect the set of constraints, modeling variables, obtained outputs, and necessary computer and human resources. A methodology for selection of an appropriate objective function for the problem of signal timing optimization was developed. The methodology was applied to a realistic case study network under four demand patterns (symmetric, asymmetric, undersaturated, oversaturated). Selection is made from a pool of five candidates: minimizing the delay, minimizing the travel time, maximizing the throughput-minus-queue, maximizing the number of completed trips (or trip maximization), and maximizing the weighted number of completed trips (or weighted trip maximization). Findings indicate that for all demand patterns, weighted trip maximization improved network performance compared with the other objective functions. Weighted trip maximization reduced system total delay by 0.1% to 5.2% in symmetric undersaturated demand, by 1.0% to 2.4% in asymmetric undersaturated demand, by 1.2% to 16.6% in symmetric oversaturated demand, and by 11.7% to 27.4% in asymmetric partially oversaturated demand. These figures indicate that the weighted trip maximization objective function is the most suitable of the candidates in oversaturated conditions, especially when demand is not symmetric. Throughput-minus-queue and trip maximization were the second most suitable objective functions for oversaturated conditions, and trip maximization was slightly more suitable when demand was asymmetric.

Arterial traffic control using reinforcement learning agents and information from adjacent intersections in the state and reward structure

An application that uses reinforcement learning (RL) agents for traffic control along an arterial under high traffic volumes is presented. RL agents were trained using Q learning and a modified version of the state representation that included information on the occupancy of the links from neighboring intersections. The proposed structure also includes a reward that considers potential blockage from downstream intersections (due to saturated conditions), as well as pressure to coordinate the signal response with the future arrival of traffic from upstream intersections. Experiments using microscopic simulation software were conducted for an arterial with 5 intersections under high conflicting volumes, and results were compared with the best settings of coordinated pre-timed phasing. Data showed lower delays and less number of stops with RL agents, as well as a more balanced distribution of the delay among all vehicles in the system. Evidence of coordinated-like behavior was found as the number of stops to traverse the 5 intersections was on average lower than 1.5, and also since the distribution of green times from all intersections was very similar. As traffic approached to capacity, however, delays with the pre-timed phasing were lower than with RL agents, but the agents produced lower maximum delay times and lower maximum number of stops per vehicle. Future research will analyze variable coefficients in the state and reward structures for the system to better cope with a wide variety of traffic volumes, including transitions from oversaturation to undersaturation and vice versa.

A Cell-Based Distributed-Coordinated Approach for Network-Level Signal Timing Optimization

This article develops an efficient methodology to optimize the timing of signalized intersections in urban street networks. Our approach distributes a network-level mixed-integer linear program (MILP) to intersection level. This distribution significantly reduces the complexity of the MILP and makes it real-time and scalable. We create coordination between MILPs to reduce the probability of finding locally optimal solutions. The formulation accounts for oversaturated conditions by using an appropriate objective function and explicit constraints on queue length. We develop a rolling-horizon solution algorithm and apply it to several case-study networks under various demand patterns. The objective function of the optimization program is to maximize intersection throughput. The comparison of the obtained solutions to an optimal solution found by a central optimization approach (whenever possible) shows a maximum of 1% gap on a number of performance measures over different conditions.

Effects of Metered Entry Volume on an Oversaturated Network with Dynamic Signal Timing

This paper analyzes the effects of different traffic metering levels at the entry points of a simulated signalized network to maintain efficient vehicle processing. Metering signals were placed along the network perimeter in advance of the bordering intersections to reduce the vehicle arrival rate and prevent oversaturation. In the simulation environment, traffic signals were externally controlled by independent agents using a learning algorithm based on approximate dynamic programming. Agents operated the signals in a cycle-free mode, reacting in real time to current demands and occupancy estimated from detectors placed at the entry and exit points of all links. The metering strategies were analyzed for delay, throughput, network congestion, and queue management. Results indicate that metering have a significant effect on network performance. Metering to levels just below the maximum throughput capacity of an intersection resulted in increased network throughput (up to 5%); reduced delay (up to 10.9%), including vehicles inside and those metered outside of the network; and queue lengths inside the network that allowed efficient use of green time. However, metering to points well below or above the capacity of an intersection did not always provide network improvements. This finding suggests that an optimal congestion level exists inside the network that can be achieved by a metering strategy. An analysis of the metering effects is presented in a case study, and field implementations and scenarios in which metering can be applied are discussed.

Does traffic metering improve network performance efficiency

Traffic metering at on-ramps in interstate highways has been widely used and led to desirable results. In urban transportation networks when demand reaches network capacity level, traffic metering may also increase network performance efficiency. In this paper, we apply different metering strategies to a case study network to see if they result in a different network operation and potentially a more efficient performance. To make sure if any observed differences in network performance efficiency is due to metering strategies and not due to an inappropriate signal timing, we determine near optimal signal timing of the network by using our Intelligent Dynamic Signal Timing Optimization Program (IDSTOP). IDSTOP incorporates Genetic Algorithms (GAs) with microscopic traffic simulation to find near-optimal signal timing parameters of the network. Our results showed that letting all traffic enter the network or metering a large portion of the traffic are not the best options. Instead metering around 20% of the traffic resulted in the best network performance in terms of average delay (16% reduction compared to no metering and 17% reduction compared to extremely heavy metering strategies), network throughput (18% increase compared to heavy metering), and average travel time (14% reduction compared to no metering and 10% reduction compared to heavy metering). Our findings suggested that in an urban network, there is an optimal point that sending more vehicles into the network than that deteriorates network performance efficiency.

A comparison of approximate dynamic programming and simple genetic algorithm for traffic control in oversaturated conditions — Case study of a simple symmetric network

The performance of two algorithms for finding traffic signal timings in a small symmetric network with oversaturated conditions was analyzed. The two algorithms include an approximate dynamic programming approach using a “post-decision” state variable (ADP) and a simple genetic algorithm (GA). Results were found by using microscopic simulation and compared based on typical measures of performance (delay, throughput, number of stops) and also on measures that considered the efficiency of green time utilization and queue occupancy of the links. The symmetric characteristics of the small network allowed a straightforward analysis of the operation of the signals, providing some insights on the quality of the solutions. Results showed that even though the solutions from ADP were very different from those in GA, the network performance for both methods was similar, used green time efficiently preventing queue backups, and served all approaches according to current demands. The potential of ADP using the “post-decision” state variable is currently under further analysis using more challenging conditions, additional constraints, and domain knowledge as part of the algorithm formulation.