Deep learning-based collision safety control system and an operation method thereof
AssigneesHyundai Motor Company, Kia Corporation, Seoul National University R&Db Foundation
InventorsHyung Wook Park, Sung Roh Yoon, Dong Hyeok Lee
Life of this patent
Estimating the term…
Abstract
A collision safety control system includes a memory storing a collision safety model having a deep learning-based collision safety control logic. The collision safety control system also includes a processor electrically connected to the memory. The processor is configured to, in accordance with the collision safety control logic, train, based on at least one signal including pre-collision data and post-collision data, the collision safety model such that the collision safety model outputs a collision type and a required time-to-fire (RTTF) of passenger protection equipment corresponding to the at least one signal.
Claims15
A collision safety control system, comprising:
- a memory configured to store a collision safety model having a deep learning-based collision safety control logic;
- and
- a processor electrically connected to the memory, the processor configured to, in accordance with the deep learning-based collision safety control logic, train, based on at least one signal including pre-collision data and post-collision data, the collision safety model such that the collision safety model outputs a collision type and a required time-to-fire (RTTF) or a time-to-fire (TTF) of passenger protection equipment corresponding to the at least one signal,
- wherein the collision safety model is trained i) by performing first supervised learning based on the RTTF as output data and ii) by performing second supervised learning based on an adjusted RTTF as the output data,
- wherein the processor is configured to obtain the adjusted RTTF by applying a weight according to a forward collision-avoidance assist (FCA) level included in the pre-collision data to the RTTF corresponding to the collision type, and
- wherein
- the collision safety model is provided in a vehicle, and
- the vehicle is configured to perform automated driving control based on the collision safety model.
A method of operation of a collision safety control system, the method comprising:
- obtaining at least one signal including pre-collision data and post-collision data;
- and
- training a collision safety model having a deep learning-based collision safety control logic such that the collision safety model outputs a collision type and a required time-to-fire (RTTF) or a time-to-fire (TTF) of passenger protection equipment corresponding to the at least one signal,
- wherein the collision safety model is trained i) by performing first supervised learning based on the RTTF as output data and ii) by performing second supervised learning based on an adjusted RTTF as the output data,
- wherein the method includes obtaining the adjusted RTTF by applying a weight according to a forward collision-avoidance assist (FCA) level included in the pre-collision data to the RTTF corresponding to the collision type, and
- wherein
- the collision safety model is provided in a vehicle, and
- the method further in includes performing automated driving control of the vehicle based on the collision safety model.
A non-transitory computer-readable storage medium having a program recorded thereon, the program, when executed by a processor, causes the processor to:
- obtain at least one signal including pre-collision data and post-collision data;
- and
- train a collision safety model for a vehicle, the collision safety model having a deep learning-based collision safety control logic such that the collision safety model outputs a collision type and a required time-to-fire (RTTF) or a time-to-fire (TTF) of passenger protection equipment which correspond to the at least one signal,
- wherein the collision safety model is trained i) by performing first supervised learning based on the RTTF as output data and ii) by performing second supervised learning based on an adjusted RTTF as the output data, wherein the adjusted RTTF is obtained by applying a weight according to a forward collision-avoidance assist (FCA) level included in the pre-collision data to the RTTF corresponding to the collision type;
- and
- perform automated driving control of the vehicle based on the collision safety model.
Dependencies are inferred from the claim text.