Tolling is a practice that has existed for many centuries, across regions. Tolls were in use as early as the European Middle Ages. The funds collected were used to support bridge construction. Reportedly, “the fee for a horse and rider was two deniers (pennies); a wagon, four deniers; a donkey, cow, or sheep, one-half denier. On Old London Bridge, tolls were charged both for traffic over and navigation under the bridge.” 
Over the years, our modes of transport have changed, the economy has changed, and with it, as have toll prices (imagine paying in pennies today!) However, tolls have remained a constant to this day. The funds collected continue to be redirected towards road maintenance but traditional tolling, as we have known it in the last few decades, is undergoing a sea of change. More factions of urban and rural population are gaining mobility. Funds procured via gas tax are declining as electronic vehicles are becoming the sustainable alternative. Manned tolls are not able to keep up with the influx of traffic. Congestions at toll booths are a regular sore sight to the eye. Majority of travel time goes into waiting at congested booths.
To maintain a steady revenue and ensure a friendly travel experience, tolling needs to become ‘smart.’ Traditional approaches need to be revamped. And at the forefront of this change stand Artificial Intelligence and Machine Learning.
AI & ML have already revolutionized the transport industry with the introduction of traffic sensors for accident detection, preempting congestion and further predicting the traffic movement to enable efficient route planning. These technologies are now transforming the world of tolling, unlocking avenues of efficiency, promoting judicious usage of time and resources, and ensuring that the vast data goldmine collected at tolls does not go underutilized.
Here are 4 ways in which AI & ML are fueling efforts that are radicalizing the tolling landscape:
Going digital is fundamental for a seamless, toll collection process. It reduces the chances of fraudulent transactions and the need to carry physical money. In the wake of the COVID-19 pandemic, social distancing has truly become the norm – and one major enabler in this regard, reducing the need for physical interaction, are cashless transactions. Moving beyond traditional tolling setups, technology is ringing in significant change in tolling, encompassing new innovations that facilitate seamless automated tolling to unmanned tolls to toll bots and more.
Here is a look at some of the popular cashless initiatives being undertaken across the world:
Radio frequency identification data or RFID data is collected through tags that are, in turn, linked to prepaid accounts. RFID transponders & readers, for instance, are a prevalent technology in tolls across the United States, India and many other countries.
When vehicles cross the sensors at a toll booth, the data collected from the RFID tag is cross-referenced to records for validation, and the account details are sent back to the toll booth. If the amount is paid, the user is let through, if not, the user has to pay the amount at the toll, or through a cashless transaction system that their toll account is mapped to. This method of data collection for tolls cuts down the need for human interaction. It also creates a repository of transactions.
While for the individual, the hassle for collecting tolls when accounting is cut down, the digital standardization of payment collection also reduces the probability for miscreants attempting to take a share of profits by altering toll rates. For the tolling officials, it helps create a repository of data that tracks movement of citizens from one location to another and running these data through models can also help predict toll booth congestion, create better toll pricing systems based on traffic movement, and so on and so forth.
Mandating the use of technologies like RFID, also helps track down possible criminal movement of nefarious activities across borders. The identification of a vehicle can also help in vehicle theft detection. When a vehicle goes missing, the owner can alert authorities and register a complaint keeping in mind the Registration ID, the RFID number, etc. Either of these aspects can be detected at the toll, enabling officials to track down the stolen vehicle.
GPS tagging tracks travel details via location tags. Toll fees are charged according to travel data collected, and only right at the end of the trip. This is still an upcoming technology. This method greatly reduces the need for tolling infrastructure. Hence, many countries are experimenting with the possibility of employing the same.
Often, even with the use of RFID, traffic jams at toll booths are unavoidable during peak hours or around holiday season, or important events. With GPS tagging, the need for toll booths is done away with. By leveraging VTS, i.e., Vehicle Tracking Systems, not only is the vehicle’s movement from one city/town/state/country border to another tracked, but the entire trajectory of the vehicle is tracked. ML models utilize this data for route optimization through assessment and mapping of appropriate neural network systems to enable swift movement.
Therefore, the data is potentially richer than the data collected at toll booths because transport authorities now have a repository that tracks movement throughout, helping AI powered ML models to better predict and analyze traffic movement to suggest remedial measures and preempt congestions. This method of toll collection is also doubly important to track goods movement or the movement of VIPs as the technology helps track the vehicle movement completely, from start to finish.
On the flipside, with GPS-tagging and similar tracking technologies, it is easier to gain information about someone's whereabouts. In the wrong hands, this might be misused, be a critical threat to privacy and perhaps even be leveraged to spy. The information gathered could also be used for illegitimate activities, putting the stakeholders involved in risk. Perhaps, the only resolve firms can ensure they undertake is to guard the information/ data collated and setup robust security systems to ensure that it is not misused.
In its existing state, across countries, video recognition software used for vehicle detection is still plagued by issues such as weather conditions, lack of visibility, etc., and the misidentified vehicle number plates, can generate faulty tolls. In this technology, facial recognition software is set up at toll booths and visual analysis helps identify drivers by cross-referencing data with government databases. Upon calculation of the toll fee, the bill is sent directly to the driver.
While this technology is still in its nascent stages, AI is bolstering its capabilities. Companies are experimenting with 'smart’ facial recognition, powered by AI, ML and 3D object tracking technology to accurately detect and identify vehicles and calculate toll fee based on the toll booth location and the vehicle’s size and capacity. This can be done seamlessly even when the weather conditions or toll booth camera's angles are not favorable.
This technology can also be leveraged to identify number plates. Machine learning algorithms swiftly recognize license plates and reduce the need for manual efforts. Pertinent data on vehicle location, lane number, time stamp, etc. can also be identified. Automated vehicle data collation also acts as a repository of vehicle traffic and behavior, can help derive details about vehicle accidents. In times of crisis, this information makes navigational support possible, and that is why beyond just tolling, this technology is pivotal for highways and roadways, in general.
Manning toll booths usually requires repetitive tasks and the upkeep of booths. One technique that enables setting up of unmanned tolls is automated video tolling. When a vehicle passes through a video toll, cameras set up in the gantry identify the number plate. Vehicle details are collected through visual analysis powered by Optical Character Recognition (OCR technology). AI then maps the data mapped to the driver's registered account, analyzing the trip details, and billing the relevant fee to the driver’s account. Alternatively, this bill can also be displayed on a touch screen and the drivers can pay the fee digitally.
Furthermore, machine learning algorithms can help unmanned toll collection systems to calculate customized toll fees, real-time in accordance with the time of day, destination details, traffic, etc. These systems can also reduce conventional efforts taken to manually determine the type of vehicle (through analysis of the number of axles it has) based on pictures taken at the toll booth. AI-powered recognition technology alternatively leverages an image sensor that can capture the picture of a passing vehicle. ML algorithms then perform visual analysis and quickly count the axles, calculating the relevant fees.
This ML model’s accuracy can be further improved over time by setting up a feedback system. Not only will this improve detection accuracy and enable monitoring, but it also helps key in updated details when a new car or truck is launched in the market and its axle numbers are not the same as its conventional predecessors.
Data collected through automated or digitally set-up tolling systems can hold the key to streamlining traffic operations. According to WHO, " The global urban population is expected to grow approximately 1.84% per year between 2015 and 2020, 1.63% per year between 2020 and 2025, and 1.44% per year between 2025 and 2030." This increase will surely be mirrored with an increase in vehicular traffic. Cities world over are preparing themselves to deal with the inevitable traffic congestion. Peak hour vehicular movement and influx during large scale events like concerts, shows, sporting matches, festival celebrations, etc. are major causes for concern. Predictive analytics applications can be merged with AI to accurately anticipate traffic behavior, based on the border movement data collected at toll booths.
Essentially, by anticipating traffic behavior, AI & ML technologies can also enable micro-tolling, i.e., tolling for general vehicle movement in city traffic. By leveraging data on traffic behavior, dynamic pricing strategies can be set up – for example, if a citizen travels to work at peak hours when the roads are congested and there is very slow vehicular movement, they are charged a higher fee in micro-tolling. However, if another citizen travels to work in times of low volume traffic, they will be charged a lower micro-tolling fee, or can even be provided perks/ earn credits, thereby motivating self-regulation from driver to driver. Alternatively, apart from time of travel, even routes can be scrutinized for micro-tolling. Taking a longer route that results in less congestion can be charged a lesser fee than a conventional route that adds to congestion.
However, while micro tolling can truly transform vehicle movement and largely reduce traffic congestion, the resources and technology needed for these systems come at the cost of added investment. Therefore, by analyzing data and predicting ROI and cost of setups, transport authorities must gauge feasibility of micro-tolling or at least prepare their budget such that it can be made affordable within a few years.
As established, smart tolling offers a plethora of unused data about vehicle movement. This unstructured data can be run through ML models to help make sense of the data and mine actionable insights. Valuable insights about high traffic hours, driving patterns, regular commuting routes, etc. can be determined. Therefore, these analyses can truly help identify which models need reinventing and ensure that tolling remains pertinent, real- time and contextualized to the geographical locales and target demographic.
These business models can enable better fuel resource management. There is a need to ensure that there are enough reserves to leverage for alternative infrastructure. One big aspect that can help balance fuel consumption is using the data collected at toll booths to track vehicular movement and identify unnecessary fuel consumption. Fuel costs are constantly fluctuating and optimizing routes to ensure that expenses are cut back on, deliveries are efficiently organized, and overall operational expenses are reduced.
Lastly, 'Smart Tolling' techniques improve quality of travel as they reduce time spent waiting in lines. It also makes traffic smooth-flowing. By truly making tolling citizen-centric, the process can be made more user-centric. Setting up a mobile application, for instance, increases ease in usage. The process can be more interactive by leveraging the use of customer service platforms. Setting up chatbots in these applications and websites can further enhance the user experience. Not only can it help resolve user queries, but it can also be used by tolling agencies to access records, compare data and perform analyses.
The world has come a long way since tolling in the middle ages. The way in which technology is advancing, even last year’s technology might become obsolete in the face of constant development. Therefore, creating dynamic tolling methods is the need of the hour. Yes, redefining traditional models to align with future technological advancements and population changes can be daunting. However, thanks to AI and ML powered technologies, this effort falls well within our capabilities.
Saurabh comes from a strong mathematical background and has over 3 years of experience in the AI & ML world with specific expertise in predictive analysis, data processing and data mining. Saurabh is extremely passionate about deep reinforcement learning, business consulting & strategic data science management using model building. Apart from work, saurabh's interests include travelling, bike riding and fishing.
Kashyap comes from an Automobile Engineering background and fittingly, is a "Petrolhead" at heart and a data scientist by profession. He is passionate about learning more about business consulting, strategic data science management and more. Apart from work, Kashyap's interests include reading about new technologies in the automobile industry and watching movies.