Browse Publications Technical Papers 2019-01-0880

Backseat Driver - Driver Advisory System 2019-01-0880

Most V2X, ADAS and autonomous driving systems today are based on the precise location and prediction of movement. These systems are computationally complex and depend on precise sensor measurements. This might not be always possible e.g. inaccurate GPS location during cloudy weather conditions. Proposed here is a new approach, the “Backseat Driver”. The “Backseat Driver” is based on heuristics and machine learning concepts for modeling driving guidance. This is very similar to how “humans” drive. The Backseat Driver complements the existing V2X and RADAR based systems by issuing an advisory to the driver. A machine learning approach (Artificial Neural Networks, Expectation-Maximization, Decision trees) is adapted in order to generate advisories. Moreover, with continuous reinforced learning, the predictions become more accurate. In the upcoming days of semi-autonomous driving, a heuristic approach for predicting and issuing advisories could make a significant difference to the driving experience. The Backseat Driver uses historical data and machine learning approach to generate advisories. Timely advisories such as merging traffic, oncoming traffic while taking a turn, busy pedestrian crossing, school/hospital zone could alert the driver to be cautious. The Backseat Driver is a low complexity, low cost system which works even if sensor data is not accurate. Unlike V2X, there is no critical number of installations required for the system to work. In contrast to V2X and ADAS systems which provide alerts only while facing a situation, the Backseat Driver can issue advisories even in advance. The current scope of research shall be limited to issuing an advisory to the driver and not in controlling any vehicle function. In conclusion, the research focus is to provide real time and advance advisory messages to the driver, without dependency on availability of accurate sensor data.


Subscribers can view annotate, and download all of SAE's content. Learn More »


Members save up to 18% off list price.
Login to see discount.
Special Offer: Download multiple Technical Papers each year? TechSelect is a cost-effective subscription option to select and download 12-100 full-text Technical Papers per year. Find more information here.
We also recommend:

Non-Invasive Real Time Error State Detection for Tractors Using Smart Phone Sensors & Machine Learning


View Details


LiDAR Based Sensor Verification


View Details


Cooperative Collision Avoidance in a Connected Vehicle Environment


View Details