Samsung's efforts in the self-driving automotive industry appear poised to continue growing, based on a recent patent for a new autonomous driver training model published by WIPO.
The patent itself applies primarily to an "apparatus" and methodology for training a self-driving vehicle's AI that utilizes machine learning and a combination of human driver metrics and traditional sensor information. The data for the former is tracked and pulled from several sources within the vehicle itself while a human is in control; that includes information such as grip strength and positioning taken from the steering wheel and brake or throttle pedal movement.
However, it also seems to include headset-gathered metrics, with Samsung explicitly indicating measurements taken via eye-tracking and a brain-wave electrocardiogram. The former of those is self-explanatory, but the latter is more closely related to and encompasses technologies more often associated with EKG, ECG or EEG measurements taken in a medical setting.
Samsung's description of the apparatus indicates that the electrocardiogram information would be used to assess the changes in a driving environment and dangerous driving circumstances in combination with those other metrics. However, it would also be compiled with the driver's use of turn signals and their "manipulation" of the vehicle's horn, stereo or other instruments in order to build a set of patterns in driving behavior to start from and improve on. That would be further underscored by metrics gathered from cameras, LiDAR, radar and navigation data in order to compile a more complete picture from which to create an autonomous driving model for the AI to work with.
Background: Although Samsung recently started filling out its portfolio in terms of AI automotive innovations, technologies and platforms, this is a relatively big step for the company. Previously, the vast majority of its ambitions could summarily be collected under the umbrella of "supportive" technologies. For example, at CES 2018 in January, the company introduced a new series of self-driving technologies that sought to make the industry more modular. To that end, the 'DRVLINE' platform encompassed both hardware and software but could be put together piecemeal and was intended for use by current OEMs in the automotive industry and service industries rather than meant for use by Samsung to create its own vehicles. For the most part, all of its technologies and press releases have centered around a similar concept, building solutions that align with the self-driving vehicle industry without taking on the tasks of building out its own subsidiary to become an active manufacturer.
Bearing that in mind, the company has also applied for and received an autonomous vehicle testing license in its home country. Specifically, that was awarded way back in mid-2017, but that doesn't mean this new patent isn't geared in the same direction. In fact, this may be among the first indications that Samsung wants to do far more than simply provide components and associated software for others in the race for AI vehicle dominance. Instead, if it puts these patents into place, the company may be preparing to compete more directly with companies such as Alphabet's Waymo, which builds its own systems for use with another manufacturer's vehicle platforms rather than selling them to the OEM.
Impact: Setting that aside, most autonomous training programs depend primarily on the use of LiDAR, radar and cameras coupled with accurate mapping data and hundreds of thousands of miles of test driving. Ordinarily, the AI is accompanied by a human driver just in case the system fails to respond or any software-related issues arise.
Samsung's concept approach is different in that it combines those with a direct real-time analysis of a human driver, going as far as reading their brain waves. While there are a lot of obvious ways that that could go horribly wrong, it may provide autonomous drivers with a much better way to handle non-autonomous vehicles sharing the roadway. Moreover, it could help improve a self-driving vehicle's “situational awareness” and improve how other unknowns in an environmental setting are responded to if Samsung chooses its human drivers responsibly and carefully.