The car controls system and its interfaces are the major witnesses of the user behaviors that ultimately lead to the emergence of safety. All of the user actions in the primary vehicle platform can be transcoded in an homogeneous XML format and logged, together with their time of occurrence. A number of other data sources can be collected and transcoded in the same format -- for instance automotive imaging, LiDAR, GPS and traffic information, V2V, and V2I data sources. This constitutes a constantly updated stream of contextual information describing the cloud of behaviors taking place within and "around" each car and each user. Said stream represents a very valuable knowledge that may be applied in several ways:
- first, it represents a knowledge base to reason about individual and social driving behaviors. Post-mortem approaches and tools can be used to correlate multiple causes leading to accidents and other safety-related conditions.
- Secondly, said knowledge base can be integrated with Advanced Driver Assistance Systems (ADAS) and used to dynamically adapt / personalize the driving experience with the objective to enhance safety, learn peculiar driving traits, compensate them with corrective adaptations in the ADAS system, and constructing a periodic personalized feedback report highlighting recursive dangerous behaviors and conditions. The result would be that of an antifragile car: a car able to systematically improve safety by learning and evolving after each and every driver.
- Third, by using dynamic profiling and machine learning it is possible to create and constantly update a stereotype of the "official" drivers. By using Hidden Markov Models, Bayesian Intelligence and similar techniques it can be possible to compare reference stereotypes with observed stereotypes. Discrepancies (see also here) imply that the either the driver has changed, which could be a possible indication of theft. Alarms could be instructed so as to automatically inform, e.g., the official drivers through their smartphones, or the nearest police offices.
- Fourth, by comparing the reference and the observed stereotypes it is possible to detect this the driving behavior is drifting away from the reference ones. By coupling this with other ADAS-relayed information, the car could "realize" that the driver is developing fatigues or is under the influence of substances affecting his/her behaviors.
Also by considering my concept from the viewpoint of marketing and advertising, the concept of an "Antifragile Car that learns how to protect You" would most definitely attract the interest of the customers.
I wonder what the automotive industries would think of such an idea.