Los Angeles’ Traffic Issues Could be Alleviated with Machine Learning

If a Los Angeles County resident is asked “what is the biggest problem facing Los Angeles?”, they typically answer with one word, traffic. Therefore, it should be no surprise that INRIX recently ranked Los Angeles the most congested city in the world. INRIX arrived at this conclusion following a global vehicular survey that spanned 1,360 cities in 38 countries.[1] It is estimated that Los Angeles’ gridlock problem costs the area over $19 billion per year and the average Los Angeles driver spends ~102 hours in congestion. Further exacerbating the issues on the road, the number of licensed vehicles in Los Angeles County has increased every year since 2010 and totaled ~8.0 million in 2017.[2] To combat the congestion, the city should consider implementing machine learning technology to optimize the process flow of vehicular traffic using adaptive technology to manage traffic signals throughout the city. Traffic light technology is antiquated; it is typically based on fixed time intervals that become ineffective during peak hours or special events (i.e. sporting events or concerts).[3] Machine learning can be implemented to manage roadways and optimize their efficiency in real time and alleviate some of these issues.

Eighty-four percent of Los Angeles commuters drove or carpooled to work in 2017.[1] This habit forced Los Angeles to be an early adopter of some advanced traffic technology. In 2000, the City implemented a transponder system to keep traffic lights green for City buses. This antiquated, yet effective system, decreased bus travel times by 25 percent.[4] Taking this concept into the 21st Century, Beverly Hills recently announced a partnership with a Pittsburgh based firm called Rapid Flow Technologies to integrate a machine learning traffic system into 15 of the City’s traffic lights. The firm’s machine learning technology system allows traffic lights, within its integrated network, to communicate with one another to clear traffic based on the volume at upstream and downstream intersections, like LA’s bus solution, but its system adapts to all vehicles. An alternative system is being developed by the Federal Highway Administration’s Exploratory Advanced Research Program and it is currently being used in Arizona and was recently approved for additional cities.[5],[6] Either of these systems could help to reduce the city’s congestion and upgrade the infrastructure to support the area’s driving culture.

In 2016, the “$120 billion Measure M ballot was approved to expand transit capacity and improve highways throughout the region… by [reducing travel time] 15% by 2057.”[7] However, Measure M does not address the demand issues that will exist after the highways are improved. Consumer behavior is shifting towards ride sharing and fighting this trend with antiquated methods is a waste of City resources. Los Angeles County should rapidly implement the machine learning technology that is being tested in Beverly Hills in the short term to verify its validity. After verification, it should fully integrate the technology into its traffic infrastructure. The traffic system in Pittsburgh show a 25, 30, and 40 percent decrease in travel time, braking, and idling, respectively.[8] If these results can be replicated in Los Angeles County, the region’s travel woes could be alleviated, and additional cost savings and reduced CO2 emissions may be realized. There is also potential for this technology to be implemented in the light rail and other mass transit systems.

Open questions remain about the validity of this proposal. Could this system reduce congestion as much as it has in other cities given Los Angeles vast highway network? The highway system does not utilize traffic lights so the effects of this system may be muted. Additionally, could any of the Measure M capital be allocated to this type of infrastructure improvement?

(792 Words)

[1] Graham Cookson, “INRIX Global Traffic Scorecard,” INRIX Research, (February 2018), INRIX Research, http://inrix.com/scorecard/, accessed November 2018.

[2] California Department of Motor Vehicles, “Department of Motor Vehicles Estimated Vehicles Registered by County for the Period of January 1 through December 31, 2017,” downloaded from CA DMV website, https://www.dmv.ca.gov/portal/wcm/connect/add5eb07-c676-40b4-98b5-8011b059260a/est_fees_pd_by_county.pdf?MOD=AJPERES, accessed November 2018.

[3] Xiaoyuan Liang, Xusheng Du, Student Member, IEEE, Guiling Wang, Member, IEEE, and Zhu Han

Fellow, IEEE, “Deep Reinforcement Learning for Traffic Light Control in Vehicular Networks,” IEEE Transactions on Vehicular Technology (Vol XX, No. XX): 1.

[4] Eric Taub, “System Lets Traffic Lights Wave Buses Through,” New York Times, September 14, 2000, https://www.nytimes.com/2000/09/14/technology/system-lets-traffic-lights-wave-buses-through.html, accessed November 2018.

[5] “City approves plan for adaptive traffic signals,” Herald Review Media, August 10, 2018,  https://www.myheraldreview.com/a/socoae/city-approves-plan-for-adaptive-traffic-signals/article_073b7216-9c6d-11e8-9ea9-1f7e16468360.html, accessed November 2018.

[6] “Next-Generation Smart Traffic Signals. RHODESNG with IntellidriveSM – The Self-Taught Traffic Control System,” Federal Highway Administration, (August 2009), https://www.fhwa.dot.gov/publications/research/ear/NextGener.cfm#future accessed November 2018.

[7] Andrea Lo, “Los Angeles’ notorious traffic problem explained in graphics,” CNN, February 27, 2018, https://www.cnn.com/2018/02/27/americas/los-angeles-traffic/index.html, accessed February 27, 2018.

[8] Jackie Snow, “This AI traffic system in Pittsburgh has reduced travel time by 25%,” Smart Cities Dive, July 20, 2017, https://www.smartcitiesdive.com/news/this-ai-traffic-system-in-pittsburgh-has-reduced-travel-time-by-25/447494/, accessed July 20, 2017.

Previous:

Darktrace: Can Artificial Intelligence lead the fight against Cyber Crime?

Next:

What does Machine Learning mean for a technology services vendor?

Student comments on Los Angeles’ Traffic Issues Could be Alleviated with Machine Learning

  1. The application of machine learning to reduce traffic and CO2 emissions is intriguing but I question how quickly it can be implemented. As you’ve pointed out, there is a clear unmet need and a potential solution where many parties benefit. It has also been implemented in smaller areas and seen initial success. However, so much of machine learning depends on training the algorithm that the complexity and scope of LA roads could pose a larger challenge. It could require many iterations over a long time to see results. I also wonder what guard rails need to be implemented to ensure safety. How does the system react to different kinds of accidents like an overturned car or a fire? How does it predict how quickly an unusual event will dissipate? While humans are not an ideal alternative, we should ensure there are appropriate controls in place before utilizing machine learning, particularly when safety is on the line. If we could accomplish that, you could be correct – there could be enormous opportunities within transportation.

  2. I found this article particularly fascinating. As you mentioned in your article, a lot of of the congestion in LA is due to the lack of capacity or resident driving behavior on the highway system. If the majority of the problems ultimately stem from this, would it make more sense to concentrate resources on applying machine learning to the highway system? Is there a way to identify where the true bottleneck is? If so, are there effective ways machine learning can be applied to the highway system? In addition, with the Olympics coming to LA in 2028, LA seems to be in a time crunch to fix the overall traffic system. With this in mind, does it make sense to dedicate more resources to this initiative or other initiatives such as expanding the public transportation or highway system?

Leave a comment