Machine Learning Approaches to Mobile Sensing Data to Make Self-Driving Cars Safer

Key Takeaways:

• Mobile sensing data from IoT devices have created opportunities for data scientists to better understand how we drive.
• Accelerometry and GPS data, for example, can be used to determine vehicular heading, acceleration, speed, climb, and other aspects of its motion.
• Machine learning and other data science techniques are allowing Dr. Michael Bell and his team at Agero to manage and interpret the large datasets of ‘mobile telematics’ in order to build systems that can help drivers during emergencies, and provide insights for automakers, insurance companies, and the logistics industry on driver behavior.

Among the most exciting trends in data science and machine learning is the growth of the Internet of Things (IoT) and an accompanying abundance of data from digital devices. In his talk at ODSC East 2018, Dr. Michael Bell, Senior Data Scientist at Agero, described how his team is using mobile sensing data collected from smartphones to help make driving safer.

Slide Copyright Michael Bell, ODSC East 2018

Over the past decade, smartphones have been packed with more and more sensors of increasing sophistication. Many of us take for granted that the simplest features of our smartphones, such as the ability to shift display from vertical to horizontal depending on how the device is being held, entail an intricate process of motion sensing and data interpretation. While many of these sensors are intended to streamline user interfaces or allow for interactive mobile gaming, Dr. Bell and his team at Agero have sought to use these sensors – and the data generated by them – to assist in the development of self-driving cars. Agero is harnessing data from mobile devices such as information about positioning and motion to better understand vehicle heading, acceleration, speed, climb, etc. Machine learning algorithms have helped Dr. Bell and his team transfer their findings into actionable insights about driver behavior.

Slide Copyright Michael Bell, ODSC East 2018

The implications of better understanding how we drive based on these mobile sensing data are significant for driver safety and the auto industry. First, Agero is working on systems that would allow phones to automatically dial 911 in the event of an accident. Second, Agero seeks to help people understand what they’re doing while driving to help individuals become safer behind the wheel. Such information could also be used in driver training and vehicular fleet management for organizations. Some experts have speculated that such information could be a game changer for the insurance industry as it constantly seeks to improve its understanding of driver behavior.

Slide Copyright Michael Bell, ODSC East 2018

Dr. Bell pointed out some of the challenges associated with mobile telematics that his team deals with on a daily basis. When it comes to using data from accelerometers and GPS to identify crashes, the system must be able to distinguish accidents from other events causing sudden movement to the phone, such as drops and bumps. The task is complicated by the fact that the data generated by phone sensors varies tremendously based on the make of the device and its operating system. Whereas iOS devices are relatively consistent in their data output, devices from the more fragmented Android market make data collection and interpretation more difficult. Nevertheless, it’s likely that the advances taking place at Agero by Dr. Bell and his team will have a significant impact on the ways that we drive in the not-too-distant future.

Interested in hearing more in-depth discussions from leaders in the data science and artificial intelligence space? Check out ODSC’s upcoming conferences in San Francisco and London and register to be in the know.