Cambridge Mobile Telematics launches next-gen AI suite for mobility

Cambridge Mobile Telematics (CMT) has launched DriveWell Atlas, a suite of telematics foundation models for AI in mobility.

Building on more than a decade of applying AI to driving data, the approach of DriveWell Atlas mirrors the foundation model revolution that is transforming industries from language to robotics, a CMT press release says.

DriveWell Atlas learns the underlying physics of force, motion, and trajectory across CMT’s platform. CMT sensor data detects distracted driving, crashes, and risky behaviors. CMT says DriveWell Atlas “fundamentally transforms how insurers and mobility providers predict, prevent, and respond to risk.”

“Foundation models are redefining what’s possible in AI,” said Hari Balakrishnan, CMT’s co-founder and chief technology officer, in the release. “At CMT, we’re applying this breakthrough to mobility. Our team has built foundation models that understand the physics of driving at an entirely new level. This means more effective risk prediction, prevention, and response for insurers, mobility providers, and the public sector, better experiences for drivers, and safer roads for everyone.”

With foundation models, a single, powerful model is trained on a massive dataset, learning the fundamental principles of that data, and can be quickly adapted for a wide variety of specific tasks, CMT says.

“Much like foundation models have transformed how AI works with language and images, CMT’s DriveWell Atlas is the first to apply this approach to the physics of mobility,” the release says. “It understands not just what happened on the road, but why. Where prior models relied on a fixed set of pre-defined risk events, DriveWell Atlas is able to discover new patterns of risk. It reveals hidden relationships in driving dynamics, enabling a new level of context, learning, and precision.

“Much like how language models understand, ‘It’s cold outside’ and ‘The temperature outdoors is low’ as the same idea, DriveWell Atlas finds equivalence and nuance in sensor data.”

CMT provides some examples of what DriveWell Atlas can detect:

• • Sudden deceleration is a defensive brake rather than inattentive driving.
  • Swerving could be to avoid a collision, not recklessness.
  • A collision was a side swipe.

“These patterns enable us to understand and respond to risk and crashes with more context and precision,” CMT says.

DriveWell Atlas will also enable CMT and its partners to uncover and validate risk factors faster; for instance, understanding how ADAS lane-keeping assistance or blind-spot warnings affect crash likelihood.

“DriveWell Atlas takes data harmonization to a new level, learning to predict missing or incomplete data and normalizing across every source,” the release says. “The result is consistent risk events and detection, no matter where the data originates. Whether from a smartphone, a Tag, a vehicle API, or an aftermarket device, DriveWell Atlas delivers the same interpretation of driver behavior.”

Images

Featured illustration provided by GamePH/iStock