Exploring AI for Road Safety and Fleet Management

Artificial Intelligence now plays a very important part in the future of traffic and transportation as governments across the globe look to create safer roads and lesser traffic congestion. We caught up with Jedrek Fulara, CTO of Sparkbit – A European software house specialised in building IoT, AI and telematics systems, to discuss disruption in the Transport Industry and the technology they will display at Gulf Traffic 2019.


Gulf Traffic: Which technologies, in your opinion, will disrupt the road safety and transportation industry? 

Jedrek: One of the fastest growing trends in transportation in general and road safety is the use of IoT technologies. Currently, the most popular solutions are those using data from various types of sensors located in both vehicles and road infrastructure. This is the basis for what is commonly known as connected cars or smart infrastructure. And I'm sure that this trend will only go further. 

The amount of available information increases steadily with each passing year, every new car model has at least a few hundred sensors installed, thanks to which we are able to accurately assess all events related to the operation of a given element (equipment or driver) - e.g. necessary repairs, driving style, the risk level generated by a particular motorist. Telematics, therefore, becomes the basis for creating detailed analysis and profiles of drivers. Having information about the travel context of a given driver (weather, road types, speed limits) as well as information on events generated by him (braking, acceleration) we can precisely determine what type of driver a given person is, what kind of risk he poses for other road participants, and which driving areas should he improve to make his travel safer.

The second very important trend in transportation and road safety is the use of artificial intelligence. Thanks to the adoption of machine learning techniques, we can analyse and evaluate not only events that have had place but also those we anticipate that will happen.  

Gulf Traffic: How does the individual driving style analysis look like?

Jedrek: To describe how the driving style analysis looks like I will use an example of the approach that we created at Sparkbit. To prepare the individual profile of a driver, we gather telematics data from various sensors. This can include GPS, accelerometer from a mobile phone or tag, OBD2 dongles or cameras, and even radars. Having the data, we add contextual information, such as road types, speed limits, real-time traffic and weather conditions from external providers. The next step is hazardous events detection – our system identifies driver’s actions, such as speeding, harsh accelerations or following another car too closely which are signs of bad driving habits.

The goal is to build the driver’s profile and evaluate each driver in terms of safe and economical driving. The collected data and results of the analysis are presented in the form of clear dashboards and reports. Our system provides the driver with feedback and tips on what he or she can improve.

Gulf Traffic: How can artificial intelligence be used to improve driving style analysis or road safety in general?

Jedrek: As I mentioned, one of the trends in improving road safety is the adoption of artificial intelligence algorithms. This is mainly based on advanced analysis of the image from the camera or car recorder. At Sparkbit, our research-lab employed AI techniques offer behaviour analysis with real-time contextual data, based on advanced image analysis. With the use of camera and advanced machine learning techniques our system can automatically detect and analyse events such as crossing on a red light, aggressive, dangerous or illegal overtaking, keeping a safe distance between vehicles, and excessive hopping between lanes. This the most advanced driving style analysis on the market allows to truly capture the real situation on the road and therefore gives drivers the most precise insight into their behaviour.

Another example of AI employment in the transportation sector is the use of video and data analytics to monitor a driver’s condition in real time. It aims to detect when the driver is feeling sleepy or distracted. Facial recognition data is collected and transferred to the cloud where it is interpreted using machine learning. These kind of systems analyse facial features in real time and determines if the driver is tired, falls asleep or is unwell. Thanks to the use of AI over time, the system can better recognise the condition of a particular driver.

Gulf Traffic: How can telematics solutions help in road accidents assistance?

Jedrek: Emergency Medical Services (EMS) are considered as the most important link in the chain of survival for time-critical car accident injuries. According to studies published in NEJM Journal Watch, the accident connected mortality rate is twice as high in counties with median EMS response times more than 12 minutes compared to those with less than 7-minute response times. Estimates show that the reduction of response times to less than 7 minutes in urban/suburban counties and less than 10 minutes in rural/wilderness areas could have prevented 13% of all car accident-related fatalities. In my opinion the only way to effectively achieve this ratio is to use the technologies based on telematics and Machine Learning. 

At Sparkbit, we created crash detection algorithm which can significantly improve emergency services response time and thus save many lives. This system gives precise detection of potential crash situations and real-time emergency notifications. What is important is the only device needed to make the system work properly is a smartphone. We implemented a model that analyses high-frequency data in real-time (100 samples per second) from an accelerometer sensor. To make the system work as accurately as possible, our machine learning model is trained on existing public data sources (NHTSA crash database) as well as on our own recorded set coming from simulations. By using these tools and methods, our crash detection algorithm can determine with high accuracy whether an accident has occurred and inform the relevant recipients in a few seconds about the need to provide help.

Gulf Traffic: What are other potential uses of telematics and AI in the transportation Sector?

Jedrek: One of the noticeable problems both in the cities of Western Europe, as well as in the EMEA region is the shrinking parking area, and thus the lack of available parking spots on public roads. Therefore, certainly one of the areas of using the potential of telematics will be to create solutions that reduce the search time for parking locations availability in the public / special areas.

Such systems will bring a lot of benefits to the drivers themselves. Thanks to parking optimisation systems, it is possible to reduce congestion in cities, unloading traffic in areas where free parking space is most often sought after. In addition, this type of system carries several environmental benefits - lower fuel consumption and combustion and limited air pollution in the cities.

Gulf Traffic: How important is AI in Traffic and Transportation for the Middle East? 

Jedrek: The Middle East is one of the fastest developing regions in the world. Thanks to its unique position, it is a great transportation link between Europe and Asia. It is also one of the biggest business and technology hubs in the world. All this means is that issues related to transportation and road safety will only gain in importance here and all solutions that will improve road situation will be intensively deployed in this region. In addition, the Middle East region grows up as an early adopter of cutting-edge telematics solutions, in areas such as connected cars and smart cities.

Gulf Traffic: What products / systems will you display at Gulf Traffic? 

Jedrek: At Gulf Traffic, we will present the driver's behaviour analysis platform that I have already mentioned, we will also show our camera-based driver behaviour analysis system which uses AI techniques. Certainly, at our stand we will also make the presentation of the crash detection algorithms we created, as well as, the parking optimisation solutions that we provide.