Traffic congestion is a major issue in cities that creates long commutes and contributes to a large amount of greenhouse gas emissions. The unpredictability of traffic patterns undermines the effectiveness of static traffic models which are unable to adapt to changes. Artificial intelligence and machine learning can be used to reduce traffic congestion by modifying the behavior of interconnected traffic signals because current traffic management systems are insufficient to tackle this growth of traffic on the road networks. First we discuss the problem that traffic congestion poses to cities and to the environment. Next we discuss previous methods that have been used in order to reduce traffic congestion as well as our approach of using artificial intelligence and machine learning to improve upon previous systems. Finally we explain the significance of our approach and the problem of traffic congestion.
Problem / Background
One of the major issues that large cities face is the problem of traffic congestion/build-up. This problem stems directly from the fact that there is an immense amount of people commuting in vehicles and public transportation, and the current traffic design system in place is ill-equipped to deal with the amount of congestion that can occur at any given time. This can lead to vehicles sitting idly at traffic lights, which in its own right contributes heavily to air pollution and greenhouse gas emissions. In fact, in an article written as recently as 2014, it was noted that the increasing severity and duration of traffic congestion have the potential to greatly increase pollutant emissions and to degrade air quality, particularly near large roadways [9]. Other detriments include increased fuel consumption, excessive delay due to time allocated by the traffic signals, and the overall decrease in safety and efficiency of the traffic. In addition to this, according to a 2009 Texas Transportation Institute report, highway travelers waste 4.8 billion hours a year stuck in congestion, which is nearly one full work week per traveler [5]. Based on wasted fuel and productivity, the overall cost reached 33 billion dollars for delays in truck operations and the total amount of fuel wasted opped 3.9 billion gallons [5].
Save your time!
We can take care of your essay
- Proper editing and formatting
- Free revision, title page, and bibliography
- Flexible prices and money-back guarantee
Previous Methods
Modern management model uses the simplest transportation management method, a static time-of-day method with minimal real-time adjustments. [1] This method only works under little variability in the traffic flow patterns and has some limitations such high latency, lack of detection of accidents, and slow reaction.
Transportation System Management and Operations (TSM&O) is one of the most frequently used traffic management systems nowadays. TSM&O has a purpose of mitigating the effects of roadway events and managing short-term traffic. TSM&O is using an Intelligent Transportation System (ITS), real-time information about highway conditions to implement control strategies and to monitor traffic. ITS control has many strategies: metering flows onto freeways, dynamically retiming traffic signals, managing traffic incidents, geometric improvements to roads & intersections (e.g. changing streets into one-way operations), and providing travelers with information about travel conditions. The TSM&O strategies are to increase traffic efficiency based on existing highways. Deploying TSM&O has shown to be highly cost-effective, but relying on this system is a limited approach because a sound base infrastructure must exist before TSM&O can be used. [10]
Advanced Traffic Management System (ATMS) is a primary subfield within Intelligent Transportation System (ITS). It includes the basic categories of: traffic control, motorist information, incident management, and institutional coordination. [2] These elements operate in concert to maintain efficient flow on the roadway system (including both freeways and major surface arterials), balance flows between available facilities, and facilitate rapid recovery from breakdowns in traffic flow which are associated with either commute or incident caused congestion. However, ATMS has a lower accuracy and poor performance under a complex real-world case.
Decision-Tree-based Green Driving Suggestion System enhances the intersection throughputs and provides drivers the best-economic driving suggestions. Signal controllers will broadcast the signal countdown message and waiting queue information to the vehicles nearby the intersection. Then on-board units can determine the best economic driving speed and provide drivers' suggestions while maintaining the max throughput within a cycle time. [3] Although it can reduce the carbon dioxide emission and boost the traffic flows, it cannot work under multi-intersections and requires all vehicles to have the same speed. Therefore, it has very limited application in the real world.
Another approach to the congestion problem is to manage the demand for highway travel. If we can include more people into fewer vehicles through utilizing public transportation, shift the time of travel through staggered work hours, and eliminate the need for travel through telecommuting, the congestion problem will greatly improve. Building more railway or bus transit systems will also achieve the same goal of reducing the amount of automobiles on highways, and thus improve traffic problems all together. The major drawback to this method is that it requires changes in the lifestyles of commuters and any individuals. [10]
Approach
In order to combat vehicle congestion and its detrimental effect, we propose the implementation of artificial intelligence in order to assist in implementing adaptive traffic lights and route suggestions, as well as to increase traffic data collection and interpretation as the push towards connected automated vehicles (CAVs) grows. Many cities already have sensors for traffic lights which activate a timing system when cars are stopped above them, however these sensors do not change lights predictively nor do they communicate between lights. We intend to utilize the infrastructure already in place by collecting data from these sensors running them through software which may process and analyze the data. The processed information serves both to train the artificial intelligence as well as to inform it of current congestion. From here, the system will control the traffic lights in such a manner as to find the most efficient method to pass all vehicles through all the lights in the least amount of time. Keeping in mind the need for swift emergency vehicle access, when a camera spots an active emergency, all lights may be turned red in order to halt all traffic in the direction of the vehicle.
In addition to controlling traffic lights, we may connect the software to street cameras and highway information controls in order to understand current traffic congestion and patterns. By introducing this information to the system, adaptive routes which optimize commute time may be dispersed through current apps, such as google maps or waze. Furthermore, the collection and analysis of this data will provide a good base upon which CAVs may build. Already having a starting point and a system with which the CAVs may communicate will augment their already anticipated beneficial effects. Improvement of traffic regulation and control is dependent upon intelligent systems and such management will control traffic more efficiently and effectively than current infrastructure. [6][7]
Before implementing controlling algorithms, however, monitoring is a must. [8] We intend to both crowd source and approach companies already involved in machine learning and artificial intelligence in order to attain software which such capabilities. Once obtained, the software will gather information and analyze patterns and correlations for at least one year (in order to account for different weather patterns) in training without the ability to control light systems or best route manipulations. After the training period, the system will first be activated in low risk settings (i.e. suburban and rural towns) and move into cities. Once established, the system may target freeways and interstates by modulating signage for best routes and additional infrastructure may be built in order to better control traffic flow on these roads.
Significance
Using machine learning and artificial intelligence to control traffic signals has the potential to reduce commute times and greenhouse gas emissions. Additionally, smart systems such as variable speed limits can help to increase safety during congestion. Studies have already shown that artificial intelligence has helped in this regard. For instance, the study in [4] which applied artificial intelligence to the streets in Lebanon found that their model had the ability to reduce queue lengths at traffic lights by 62.82% and queuing time by 56.37%.
Figure 1. U.S. Greenhouse Gas Emissions by Sector [5]
Using adaptive traffic control signaling also has the ability to reduce greenhouse gas emissions. Referenced by [5] and according to the U.S. Environmental Protection Agency (shown in Figure 1), the transportation sector is responsible for 28% of all emissions, and is tied for first with electricity production. Also, according to [5], “93% of the energy consumed in transportation currently comes from fossil fuels.” Traffic congestion leads to an increase in greenhouse gas emissions because traffic that is idling is not moving and only releasing gases. Additionally [5] also mentions other problems associated with congestion: “travel delays due to traffic congestion cause drivers to waste fuel, increase air and noise pollution, and be stuck in their cars for extra time, leading to lost productivity.” In conclusion, adaptive traffic control systems can be extremely significant in reducing traffic congestion as well as greenhouse gas emissions.
Stuck on your essay?
