Autonomous driving system and method of controlling same

A method of controlling an autonomous driving system, the method comprising:

• generating first learning data;
• generating second learning data corresponding to a lower layer of the first learning data by performing machine learning of the first learning data using a machine learning algorithm by applying driving environment variables of an autonomous vehicle thereto;
• and
• updating the first learning data from the second learning data when the second learning data meets a predetermined first condition,
• wherein the updating of the first learning data comprises identifying that two or more sibling nodes within the second learning data share common learning information, so as to determine that the predetermined first condition of the second learning data is met, and updating a first parent node of the identified two or more sibling nodes in the first learning data, independently of other parent-child node subsets in the first and second learning data, and
• wherein the updating of the first learning data further comprises updating the second learning data from the first learning data, including the first parent node, when the first learning data meets a predetermined second condition, such that the updating of the first learning data and the updating of the second learning data are performed bidirectionally, and
• wherein the method further comprises controlling the autonomous vehicle according to an autonomous driving algorithm using the updated first and second learning data.

An autonomous driving system comprises:

• a learning device generating root learning data by performing machine learning for raw data using a machine learning algorithm and generating a plurality of first layer learning data corresponding to a lower layer of the root learning data by performing machine learning of the generated root learning data by applying driving environment variables of an autonomous vehicle thereto;
• and
• a learning control device controlling the learning performed by the learning device and updating the root learning data from the plurality of first layer learning data when a first integration condition for the plurality of first layer learning data is met,
• wherein the learning device generates the plurality of first layer learning data by applying a first-level driving environment variable among the driving environment variables of the autonomous vehicle during machine learning of the root learning data,
• wherein the learning device generates a plurality of second layer learning data by applying a second-level driving environment variable during machine learning of at least one of the first layer learning data,
• wherein the learning control device identifies that two or more sibling nodes within the second layer learning data share common learning information, so as to determine that a second integration condition of the second layer learning data is met, and updates a first parent node of the identified two or more sibling nodes in the first layer learning data, independently of other parent-child node subsets in the first and second layer learning data,
• wherein the learning control device identifies that the plurality of first layer learning data share common learning information, so as to determine that the first integration condition of the first layer learning data is met, and updates the root learning data from the plurality of first layer learning data including the first parent node,
• wherein the learning control device updates the plurality of first layer learning data, including the first parent node, from the root learning data when the root learning data meets a propagation condition, such that the updating of the root learning data and the updating of the plurality of first layer learning data are performed bidirectionally, and
• wherein the autonomous vehicle is controlled according to an autonomous driving algorithm using the updated root learning data and the updated first layer learning data.