Worldwide pioneering online software identifies road risk on a section of the Central Highway
Nearly 1.3 million people die every year on roads worldwide, while between 20 and 50 million suffer non-fatal injuries, according to WHO studies. A team of researchers from the Complex Engineering Systems Institute (ISCI) together with Autopista Central accepted the challenge of creating a sophisticated model that identifies accident patterns, sending real-time alerts when favorable conditions permit, generating a tool with potential impact on the lives of drivers and passengers in Chile and around the world.
Implemented on Santiago’s Central Highway in 2018, the innovative Road Accident Predictive System (SPAV, in its Spanish initials) is capable of generating early warnings of road accidents. The project came into existence in 2015 with the aid of data mining tools to identify under what conditions accidents occurred in the past and under what conditions the vehicles circulated normally. This, thanks to precise information collected through sensors and collection gates of the Highway, which measure flow (the number of cars that pass through the electronic toll collection gates) according to type of vehicle (light, heavy, motorcycle), average speed, and density (number of cars/unit of space).
From the relatively stable patterns of traffic accidents, the researchers created a model based on machine learning tools. It is constantly trained in order to learn to recognize patterns and since it receives information in real time, is able to calculate if there is a greater risk than usual for an accident to occur over the course of the following minutes.
The model learns and is constantly improving. Each new day of information allows it to make better decisions by identifying potential changes in accident patterns. Thus far, the tool has managed to predict 80% of the events in a specific section at a specific time.
Franco Basso, project researcher, explains “It’s interesting that some patterns are counter-intuitive. One would tend to think that accidents occur when driving extremely fast, however, it is more likely to suffer a rear-end collision when there is a speed differential. If for example there is congestion that opens up by a few meters, drivers tend to increase the speed to make up for lost time, thus increasing the probability of an accident if a short distance along the road they unexpectedly come across cars moving at abnormally low speeds of under 60 km/hr.”.
The research team is currently integrating neuroscience and artificial intelligence tools in order to determine how to intervene on the Highway and create a scenario that improves road safety. Lab tests showed that certain light frequencies can alert drivers subconsciously and lead them to assume a more cautious attitude in situations of risk, without distracting them from the actual driving. In a second stage, we will work with a state-of-the-art eye-tracking technology that follows the driver’s eye movement and measures to what extent it can be affected by different types of visual stimuli from the luminous signs and portals along the Highway and how their level of alertness changes.
“We are optimistic about the potential of the Road Accident Predictive System. The integration of tools from other disciplines has allowed us to ask ourselves more creative questions. We have reached new barriers and we can set more ambitious challenges that will ultimately increase the potential safety of people on roads around the world.”, concludes Franco Basso.