Developing safety metrics for automatic vehicle parking using machine learning

Human depth perception requires binocular vision, therefore an impairment in a single eye can eliminate or drastically diminish depth perception. When such vision impairment occurs later in life, the brain relearns the calculation of depth perception based on prior experiences, however for vision loss occurring early in life, there is no catalog of information for the brain to use to draw conclusions about depth, and the person develops a type of "learned perception". This is greatly hindered when gauging depth while driving an automotive vehicle, especially for the close-range depth perception necessary to park a vehicle near an obstacle. The use of rear-facing sensors for vehicles has become commonplace but front-facing sensor options to assist with forward-facing distance measurement have not. In this paper, a front-facing distance sensor system for vehicle parking obstacle has been proposed and evaluated for obstacle avoidance during vehicle parking for assessing and enhancing safety for automatic and semi-automatic parking system and for developing safety metrics (the safety metrology) for enhanced safety in vehicle parking. In addition, the sensor data have been proposed to be processed with machine learning algorithms (e.g., the support vector machine, SVM) for developing machine learning-based advanced control strategies for automatic and semi-automatic parking system for obstacle avoidance during vehicle parking. The results as a whole can help drivers predict obstacles during parking and park with varied levels of safety.