Ciencias Exactas y Ciencias de la Salud

A study developed by the University of West Virginia analyzed the vehicle collisions, these occur due to the slow reaction time (RT) of humans. The study involved human RT under specific conditions, they found out that fully aware drivers have an estimated RT between 0.70 to 0.075 seconds, unexpected but normal situations like a lead car brake’s lights, is 1.25 seconds, and for surprising events is estimated to be around 1.50 seconds. Therefore, the presented work provides a solution to implement an Advanced Driver Assistant System (ADAS) level 1 called Emergency Brake Driver Assistant System based on Traffic Lights (EBDASTL) using a Jetson TX2 and a ZED Stereo camera to detect Traffic Light States (TLSs), estimate the distance to a Traffic Light (TL), and perform a brake decision based on the TLS and TLD that can have a better response time than human RT in surprising events. The main contribution of this research project is the implementation of a single ADAS that has three stages. The Traffic Light State Detection Model (TLSDM) stage using EfficientDet D0. The Traffic Light Distance (TLD) stage using a ZED Stereo camera, and the Traffic Light Decision-Making (TLDM) stage using Fuzzy Logic. Up to date there is not a related work that have the three stages. The second main contribution is the on Road test performed in Queretaro Mexico, where all the components of the EBDASTL have been mounted in a car and tested in a real-world scenario. The experiment consisted of detecting red and green TLSs at six different positions (5, 7, 9, 11, 13, and 15 meters from the TL). The TLSDM achieved a mean Average Precision of 96% for distances lower than 13 meters, and 89.50% for 15 meters. The TLD achieved an overall Root Mean Squared Error (RMSE) of 0.84 for all distances. The TLDM provided a smooth brake profile. Finally, the EBDASTL provided a response time of 0.23 seconds.

Ciencias Exactas y Ciencias de la Salud

Browse

Filters

Settings

Sort By

Results per page

Search Results