Every driver and fleet safety vendor these days are claiming that they use AI, machine learning, and computer vision to enable superior solutions for their customers. But how smart is their AI?
A vendor’s ability to detect risk and reduce collisions depends on it’s in-vehicle AI technology models and camera processing power combined with sophisticated AI and data science in the Cloud. So before you build or select your AI fleet safety system, make sure you ask these 5 simple questions:
1. How many AI-powered miles are fueling their neural network?
Once AI can understand driver behavior, traffic elements, and vehicle movement risks as a human coach would, it then needs to be trained to recognize the same behaviors across vastly varying driving ecosystems, including driver characteristics, attire, cabin sizes, lighting conditions, road conditions, traffic patterns, and more.
To successfully recognize patterns and insights in the complex transportation ecosystem, one must develop a deep learning neural network, or a set of algorithms designed to recognize patterns with precision and speeds similar to the human brain—but in an automated, efficient manner. Similar to the human eye and brain, AI will only be able to account for such varying conditions if it is continuously trained with new real-world driving data. So when considering a vendor’s solution, you need to ask how many AI-powered video miles have they analyzed to be able to accurately interpret what is happening in the interior cabin, the exterior road ahead, and to the vehicle itself.
2. What AI technology model are they using?
There are many ways to implement AI—for example, on the network edge (in this case, in the vehicle), in the cloud, or end-to-end from the edge to the cloud. Most video telematics and dash cam solutions today require driver video to be uploaded to the cloud for analysis (i.e. for human review) before any high-risk event determination is made. There are significant shortcomings to this approach because of the time lag, or latency, resulting from data transmission from the vehicle to the cloud and back again—which delays real-time alerting. The latency within the cloud is 3x longer in duration when compared to a request processed on the edge.
While edge AI processing is purpose-built for real-time collision avoidance systems, cloud-native software enables rapid iteration for model improvement and offers advantages of high availability, scalability, and reliability. To make the best use of hardware resources on the market today, AI-based driver safety solutions should be implemented across both the edge and the cloud, taking advantage of what each delivers (and does best), allowing you to predict, prevent, and reduce the occurrence of high-risk events in highly complex driving environments before they happen.
3. What can their AI detect and predict in real-time?
Some cameras apply AI to detect events happening in front of the vehicle such as tailgating while others analyze certain driver behaviors inside the vehicle such as holding a cellphone. However, to help drivers avoid as many collisions as possible, you need a system that analyzes what is happening both outside and inside the vehicle, at the same time, in real-time, and synchronizes the AI model output to provide timely alerts.
For example, a distracted driver needs to be alerted sooner to have enough time to react and avoid a collision. An AI system that only analyzes exterior road events will not provide a timely alert. Further, in order to avoid as many collisions as possible, a driver needs to be warned differently depending on whether they are at risk of hitting a vehicle, a pedestrian, a bicycle, or other object. Finally, does the AI detect the highest risk events like driver drowsiness and fatigue? Also, when analyzing events, is it looking at a few frames or is it able to process hundreds of frames to understand the evolution of behavior over time? What happens if the driver’s eyes are not visible - will detection still work? Make sure you understand the full range of what the AI is capable of detecting to reduce driving risk as much as possible. It can mean the difference between saving lives and saving millions of dollars vs. suffering terrible losses.
4. Can their AI deliver precision, clarity, and real-time response?
When it comes to preventing injuries and saving lives, you want to make sure you have an AI fleet safety system that has the highest rate of precision in detecting events inside and outside of the vehicle. Who wants a safety system that only works 60-80% of the time? In order to truly reduce risk and save lives, your AI fleet safety solution must deliver greater than 90% precision, provide clear video with a large field of vision (not just the vehicle ahead), and leverage AI to simultaneously assess real-time risk from driver behavior, traffic elements, vehicle movement, and critical contextual data.
5. Does their AI architecture have enough processing power to reduce risk?
In order to reduce risk and provide drivers with real-time alerts and more reaction time before events occur, the AI architecture (both hardware and software) must have superior image processing capacity, memory, and speed to run deep learning algorithms on the edge for greater precision in interpreting complex video scenarios. Many ADAS and video telematics solutions were not developed or intended to provide real-time AI, so they do not have the right AI architecture or processing power to truly reduce risk and proactively improve driver behavior.
Smart AI fleet safety systems are purpose-built to help fleets become safer and smarter, keep drivers safe behind the wheel, vehicles on the road, and liability claims at a minimum.
Learn more about AI Fleet Safety and Risk Reduction Solutions at Geotab Extend, March 23-24, 2021
Geotab Extend is the perfect backdrop to gain a better understanding of the AI technologies that are powering fleet safety solutions today. Register today to check out Nauto’s virtual booth or drop by the Nauto “How Smart is Your AI?” Networking Room to learn why the Nauto Driver and Fleet Safety Platform is the smartest AI solution with:
- 1 billion+ AI-analyzed video miles for continuous learning
- The highest rate of precision, clarity, and real-time response
- 5-20x more processing and compute power than competing vendors
- Proven track record to reduce risk and improve behavior in 4 out of 5 drivers (without manager intervention) within 2 weeks*
*Driver Safety Study. Nauto (2019)