••• Welcome to Transport Tech Site, your one-stop-site for all the news and brand new research on tech, AI, and transportation, delivered to you as the tried-and-true blog post. From cars to planes and trains to ships, we cover it all and give an in-depth look of how humans, and human safety, grows with this ever growing field of new people, ideas, and products in the world of transportation. Come along with me as we explore these far reaches of transportation and tech while also giving you a better insight into this vast, interesting world. And, keep your minds open as I seek to answer what some expectations or pieces of information which the general public should be knowledgeable about regarding these new technologies in transportation. Along the way, I hope you can learn a few things as well and educate yourself on false conceptions on AI today, and hold new ideas in your mind on the future of us and technology. •••
An Overview of: Machine Learning, Safety, and Transportation Today
Cruises, Cargo Ships, and Safety
Have you ever thought about the sheer amount of effort which goes into transporting goods from Point A to Point B from across the globe? The seemingly simple, almost mundane task is given little to no thought by someone when they buy a product overseas. And, as it turns out, far more happens than what one might expect. In fact, "over 80% of global transportation goods are conducted by ship", a staggering amount by any regards, which is why it is important to share and discuss (Cheng et. al section 1). For this section, we will focus on what goes into keeping cargo and cruise ships running and safe.
One very exciting development in maritime safety systems is the development of better crash or collision prediction algorithms and software. This is outlined neatly in Zhaoxi Cheng's and group's research paper Multi-ship Encounter Situation Analysis with the Integration of Elliptical Ship Domains and Velocity Obstacles (a mouthful for sure!) on their own model of modeling a ship's position, speed, and the direction it is traveling in. They identified that previous models did not account for how ships would likely move and traverse the water through just simply having a 'circle of danger' for where another ship would be in there would be a likely collision. Instead, they used an elliptical zone of influence and consequently ran simulations of different scenarios with different numbers of vessels with different parameters. One major improvement they made was that they kept track of the vessel's speed as it changed over time, instead of simply taking just one reading of it all. Through this, they saw if they could accurately predict the fastest velocity the ships could go at until a collision would be imminent. They also predicted whether the ship, if it kept its current velocity, would crash with another one, after an 'encounter' with the ship. While their results were generally successful, more research must be done before it can be implemented more widely. However, they are hoping to make it easier, safer, and more accurate for ship's captains and those running the ships in riskier, more complex situations.
Another side to ships, ship safety, and the potential for collisions comes in the form of human-with-human behavior. This happens especially in emergency situations, of which multiple different kinds at sea could occur, from overboard people to medical emergencies to power failure within a large cruise vessel. These lesser studied events, while they may seem less common and thus less impactful, are just as important as the day-to-day routine of everyone onboard a larger-sized ship. This is detailed well in Andreadakis' paper in where a general analysis is given on current efforts on modeling and predicting human behavior during maritime evacuations: Investigating Abandonment Errors in Cruise/Passenger Ships: Researching the Reasons Leading to Life-Losses During an Evacuation. Using a qualitative approach to understand the "causes behind the accidents and how these can affect the abandonment process", they discuss how "the human element" and how ship design impact the evacuation process in large passenger ships (Andreadkis et. al abstract). In short, they concluded that even though current models on modeling evacuation routes in specific ship and floor layouts were good, there is a harsh ceiling to how well machine can model unpredictable, chaotic human behavior on the individual level, especially during a unique circumstance like this. However, the most important idea to take away from this is to understand that AI has not come close enough yet to recreate human behavior, being able to "represent only basic human behavioral norms" as its inputs (Andreadkis et. al section 3.1). On the other hand, there are still a very wide variety of uses for machine learning which extend even farther in the transportation industry.
Railway Transportation and Safety
The next specific area of interest I would like to discuss is the railway transportation sector. This sector, while smaller in size in its impact on the international transportation of goods, still plays a fundamental role within our world economy. To narrow our focus even further, consider the differences between which someone operating a train and someone operating a cargo ship would have to consider safety-wise. One major difference would be the terrain and land itself, with a greater inability to see and communicate in numerous places while on a train. With these greater challenges in mind, especially for rural areas and ones where communicating via an already established communication service is impossible, one group led by Aida Eduard proposed a clever solution. Using a portable receiver/transmission setup, they successfully demonstrated a way to communicate between train conductors and rail line workers to warn workers of incoming trains or stations on estimated arrival times (see fig 1 below). This was done with the help of machine learning, of course, to estimate distances between them and the train from the radio signal between, reading the signal output and strength, without using external navigation help.
Figure 1: A simplified diagram of the worker/rail system
While successful, more work has to be done to ensure the accuracy of the system is high enough, the system is configurable enough for multiple kinds of trains and areas and regions, and most importantly, more research needs to be done on the topic. Another adjacent problem which perhaps addresses a different issue is the rail infrastructure itself, and the impacts it has on the trains that ride on it on a daily basis to the environment. While not directly related to the rise of AI, I feel as if this is still important as the railway infrastructure we have today all throughout the world is beginning to deteriorate, and there is not a clear method to replace or significantly improve large swaths of rail line in an environmentally friendly way. So, I have linked below a paper which explains the issue well and in-depth of the path towards a less carbon-intensive and lower-maintenance trackbed. I think it's pretty neat.
A U.S. Airline Safety Review
To wrap up our exploration of modern transportation technology and safety, I would like to look at a commonly misunderstood and overexaggerated area of concern, commercial flying. While not dismissing any legitimate psychological conditions or personal experiences one may have had, I would like to quickly point out the facts on flying today. The NTSB (National Transportation Safety Board) released a report on commercial airline safety incidents from the past decade, and what is shown below seems to be very promising.
Figure 2: A diagram on the left of the total amount of airline accidents reported by the NTSB from 2012 to 2021 as a line graph, and the accident type as a bar chart on the right.
As seen in fig 2, the total amount of accidents per 100,000 flight hours, while already small, is decreasing, and from those accidents only a small percentage of them are fatal. There are also a wide variety of causes with the majority of which being human error. This also surprisingly doesn't have 'a small bit of turbulence' as a leading cause for airline accidents. Case in point, technology has gotten to a point where the average person does not need to worry about their safety in the hands of someone else while on many forms of transportation, which is a testament to the years of successful changes and new implementations made by organizations like the NTSB and the culmination of human effort and collaboration. And just remember to take a deep breath and think whenever our imagination and what we hear online get the better of us.
Autonomous Cars, Car Safety, and Humanity's Future
Progress on Autonomous Technology in Cars Today
While it is touted by many that traveling by plane is much safer than traveling by car, many steps are being made to ensure the safety of passengers and drivers alike with the use of autonomous vehicles on the road. Let's take a quick look into the progress we have made so far in the world of self-driving cars and car technology.
Even while many may argue to the extent to which self-driving cars will have on the industry as a whole in, say, 10 to 15 years, it is undeniable the presence self-driving cars have on our cities' transportation network today. You may have even seen those kinds of cars driving through populous city centers and next to airports with what looks to be a dome and maybe some cameras on its roof. While still impressive, most cars you see of this type are limited in where they can go, mostly confined to well-mapped and well-tested routes. What most people think about as a car having full self-driving capabilities are still farther away. But, as stated confidently in a Forbes article by Naveen Joshi, as the technology continuously evolves, "self-driving vehicles will gain mass confidence and become mainstream in the consumer realm." (Joshi para. 8).
Figure 4: A YouTube video of a review of Tesla's self-driving capabilities
Video caption: AI DRIVR. “Tesla FSD V12 First Drives (Highlights).” www.youtube.com, 20 Feb. 2024, www.youtube.com/watch?v=mBVeMexIjkw.
A nice example of the showcasing of this currently developing technology comes from Tesla's self-driving capabilities within its cars (for a fee, of course). The video above (see fig 3) and the channel connected to it provide insightful comments and critique on the latest version of this self-driving software package, and give a good, unbiased view into what it is exactly like to be in one of those things. While it may seem scary at first, it is still possible for you to take control of the car at any point you would like to. An interesting issue that comes up as well when discussing self-driving cars on the road is the problem of computers taking a much safer and slower approach to driving as compared to humans. This is seen best in how self-driving cars adhere to the speed limit very strictly and usually assuming the safest option when outside human detection/interaction is factored in. Even through all of the quirks which such a new technology must face, I believe it highlights well the strange yet exciting time we live in right now and the potential for what is to come for humanity and the future of transportation.
Roadside and Crash Safety Improvements
After one has thoroughly considered the technology that goes into making the car itself, one must consider the roadside and traffic structures which it must interact with and the possibilities still for danger out on the road. Vadim Struk of Relevant put it best by claiming correctly that "intelligent systems can predict traffic patterns and adjust schedules accordingly", which in turn "helps companies minimize delays and increase the throughput of transportation networks", which are even more uses for AI on the road than what it seems like at first glance (Struk para. 13). One seemingly niche yet still important application of this is the nature of car crashes, more specifically hit-and-run crashes. Liang Xu and his team in their research paper Identifying nonlinear effects of factors on hit-and-run crashes using interpretable machine learning techniques look at the different factors for which hit-and-run crashes correlate with or even factors which they may arise from. They trained a predictive model to predict and identify the location and type of hit-and-run accidents and to extrapolate useful data from the scene itself.
Figure 4: A plot showing correlation data between pairs of potential influencing factors of hit-and-run crashes
As seen above in fig 4, all of the factors which the group considered and how strongly they correlate with each other are shown, with blue being a higher correlation and red being lower. This is in essence saying that if we are given a certain information about or surrounding the incident, how likely would it be that we could correctly guess more information about the scene. This is impressive because it shows, through their success in training a Machine Learning predictive model as what they conclude in their report, the ability to model more complex and non-linear situations more accurately than ever before. Basically, a model being nonlinear is simply saying that it takes its different inputs into account by different amounts, mapping and modeling more realistic and human-like behavior in this case. Thus, a compelling argument can be made that the future for humanity is shaping out to be an exciting one with rapid developments happening every day all around us, whether we realize it or not. I am excited, and I hope that you are too, of what there is in store for us in transportation, transportation safety, and transportation technology in the not-to-distant future.
Reflection
For my project, my purpose was almost to entirely inform and educate my audience with current and relevant information on transportation technology. However, I tried to add in a few elements of persuasion as well when dealing with common misconceptions about a topic and to entertain occasionally with some word choices and 'blurbs' to add an ever so slightly less formal and more approachable tone. I made this decision because I felt this was the best way to introduce and explain others to this topic of interest and I felt it fit my target audience well, who would be anyone who comes across my blog and with any previous interest in technology as a whole. Such people I would consider to be closer to the level of the general public than expert researchers, but anyone between them on a scale would be an alright fit. The genre which I chose was the scientific blog post because I felt like it aligned the best with the topics I was researching on and was interested in, while allowing for some creative headroom which a full-on science report simply does not have. I do not believe I broke too many genre conventions with my choice of font and how I styled the webpage itself, and also the transitions between topics with headers and sub-headers. However, strictly speaking, I do not have all of the additional 'fluff' which comes with full-on dedicated websites for a specific blog or corporation, as in the extra or 'recommended' blog posts/articles at the bottom or the ability to switch pages to different topics, which would have been quite difficult for me to have done.
The two rhetorical appeals which I consciously considered when writing this were kairos and ethos, kairos, of course, for keeping track of up-to-date information and to appeal to the general public/informed person on events and current issues we are experiencing now in 2024, and ethos to build up credibility. Specifically, I used ethos in my choice of choosing more research papers than popular credible sources to lean more heavily on the data/analytical side of presenting the information, and, when necessary, to "raise higher" and use more positive, formal language when talking about a certain person or group of people directly. While I felt like a heavier use of rhetorical devices were less called for in my genre, I tried to make a few allusions to broader topics and to even within my other blog posts themselves for a more interesting, cohesive writing style. I tried with my tone in general to sound somewhat professional, which whether I like it or not comes from the amount of jargon and specific language used when referring to something as specific as transportation technology. What I tried to use the most and to use as varied as I could was the usage of media, specifically images and data from research papers, videos, and generic stock photo images to "round out" the blog style better, as the writing itself is only a part of the complete picture. I also wanted to add visual interest to the writing as well and better break up longer sections of text which would otherwise be harder to follow.
For research, I emphasized my search towards recent scholarly research papers with also some, but less popular credible sources as well, as I felt like it would best accomplish my goal to inform the public. I also felt like it was the most conducive to tricking my brain almost into a sense of discovery and joy whenever I came across a good research paper, as the topic is more than niche and new enough to have only been looked less than a hundred times and cited less than twice in total, so it was a fairly fun yet quite broad search. The biggest roadblock in citing sources was the shift I had to make towards research papers and away from novels, texts, and literature in book form which I was quite familiar with by the end of high school last year, and the MLA conventions for titles, lack of page numbers, etc. But beyond that, I did not experience too much trouble.
For the drafting process, I can claim I was not very efficient as most of my work were done in relatively quick, very spaced out spurts of creativity and writing, but it is the way I work the best, I find out. However, what really helped was planning out the entire structure of the website first, even though I knew I wouldn't keep each and every thing I put out initially, as to have something for my brain to latch onto when I felt more stuck in a rut mentally. Even though I did not experiment with this, I felt like AI (ironically) could have helped with the process and could have had that same role of coming up with initial ideas, and I would be curious as to see if anyone else had success that way. I did not have a plan or set schedule for when I drafted this, although I preferred more quiet places with maybe a few people around at once so as not to zone out consistently, and I could not draft with any other external noise/distractions in the background even if it was difficult to do so. One thing that I may could have done differently would've been to allocate a couple hours or so specifically for the drafting process beforehand and to reward myself afterwards to make sure I actually do it, as I know procrastination is a common problem among the class.
I learned from the feedback process that there were many opportunities for me to break up and simplify my sentences to potentially make it more digestible for my readers. This was especially true with my first sentences of each of my subsections, as I had in my rough drafts longer and more convoluted word choice and sentence structure. While I naturally tend towards laying out most of my writing in such a way, I tried to be more conscious of that throughout my writing as to make it more approachable to others. And, one final thing I noticed when doing the feedback process myself on other's papers was the variety of genre types and the different ways they were all formatted, with even a few other websites thrown in there as well, making me feel better about my decision to get out of my comfort zone making a website/webpage in a creator I had no experience in. I had a lot of fun throughout the entire process and I would definitely recommend challenging yourself with MP3 to get the most satisfaction out of it.
Works Cited
Andreadakis, Antonios, et al. “Investigating Abandonment Errors in Cruise/Passenger Ships: Researching the Reasons Leading to Life-Losses during an Evacuation.” TransNav, the International Journal on Marine Navigation and Safety of Sea Transportation, vol. 18, no. 1, 2024, pp. 221–228, https://doi.org/10.12716/1001.18.01.23.
Cheng, Zhaoxi, et al. “Multi-Ship Encounter Situation Analysis with the Integration of Elliptical Ship Domains and Velocity Obstacles.” TransNav: International Journal on Marine Navigation and Safety of Sea Transportation, vol. 17, no. 4, 1 Jan. 2023, pp. 895–902, https://doi.org/10.12716/1001.17.04.16.
Eduard, Aida, et al. “Ad-Hoc Train-Arrival Notification System for Railway Safety in Remote Areas.” Internet of Things, 1 Jan. 2024, https://doi.org/10.1016/j.iot.2024.101062
Joshi, Naveen. “How AI Can Transform the Transportation Industry.” Forbes, 26 July 2019, www.forbes.com/sites/cognitiveworld/2019/07/26/how-ai-can-transform-the-transportation-industry/?sh=31dd17ba4964.
National Transportation Safety Board. “General Aviation Accident Dashboard: 2012-2021.” Www.ntsb.gov, www.ntsb.gov/safety/data/Pages/GeneralAviationDashboard.aspx.
Struk, Vadim. “AI in Transportation Explained: Benefits and Applications.” Relevant Software, 1 Jan. 2024, https://relevant.software/blog/ai-in-transportation/.
Xu, Liang, et al. “Identifying Nonlinear Effects of Factors on Hit-And-Run Crashes Using Interpretable Machine Learning Techniques.” Journal of Transportation Safety & Security, 6 Jan. 2024, pp. 1–21, https://doi.org/10.1080/19439962.2023.2299005.
Stock Image Credits (From Top to Bottom)
Adobe. “Concept of Metro Railway System Engineering Infrastructure,” Adobe Stock, stock.adobe.com/images/concept-of-metro-railway-system-engineering-infrastucture/383182315?asset_id=383182315.
Adobe. “Smart Logistics and Warehouse Technology Concept, Real Time Location Data Tracking Freight Shipment Delivery, Container Ship at Port, Global Business Logistics Import Export Transportation Background,” Adobe Stock.
Adobe. “Self Driving Electric Car without Driver on a Busy Street. Autonomous Mode. Head-up Display.,” Adobe Stock, stock.adobe.com/images/self-driving-electric-car-without-driver-on-a-city-street-autonomous-mode-head-up-display/228815038.