Method for determining safety-critical output values by way of a data analysis device for a technical entity
View Patent ↗A method for determining safety-critical output values by way of a data analysis device for a technical entity. The method includes receiving data and/or measured values for the entity by way of the data analysis device. The method further includes processing the input values (x) by way of the data analysis device using a software application in order to determine at least one first output value (y 1 ). The method further includes using a neural network having a plurality of layers (h θ (x)) with first learnable parameters (θ); modifying a layer of the neural network using a function (φ) so that the output value (y 1 ) is located within a defined value range (C(s)) of at least one target parameter (s); and determining the target parameter (s) and/or the value range (C(s)) of the target parameter (s) using further additional layers (k β (x)) of the neural network with second learnable parameters (β).
1. A method for determining safety-critical output values (y 1 , y 2 , . . . , y n ) by way of a data analysis device for a technical entity, said method comprising:
receiving data and/or measured values for the entity and/or surroundings of the entity by way of the data analysis device, wherein the data/measured values describe at least one state and/or at least one feature of the entity and/or the surroundings of the entity and constitute input values (x);
processing the input values (x) by way of the data analysis device using a software application in order to determine at least one first output value (y 1 ), said processing step comprising the following substeps:
(i) using a neural network having a plurality of layers (h θ (x)) with first learnable parameters (θ);
(ii) modifying the last layer or an additional layer of the neural network using a function (φ) so that at least one first output value (y 1 ) of said output values (y 1 , y 2 , . . . , y n ) is located within a defined value range (C(s)) of at least one target parameter (s); and
(iii) determining the target parameter (s) and/or the value range (C(s)) of the target parameter (s) using further additional layers (k β (x)) of the neural network with second learnable parameters (β).
2. The method as claimed in claim 1 , wherein the target parameter (s) and/or the value range (C(s)) constitute at least one second output value (y 2 ) of said output values (y 1 , y 2 , . . . , y n ).
3. The method as claimed in claim 1 , further comprising passing a representation (g(s)) formed from the target parameter (s) onto the neural network as a further input value such that a calculation value (z) of the neural network also depends on the target parameter (s).
4. The method as claimed in claim 1 , further comprising adding a noise component (Δs) to the determined target parameter (s), the noise component (Δs) being a random number with the same dimension as the parameter (s).
5. The method as claimed in claim 2 , further comprising jointly or separately using learnable parameters (γ) to determine the first output value (y 1 ) and the second output value (y 2 ).
6. The method as claimed in claim 1 , further comprising using a database that stores data relating to properties of the technical entity, images and characteristic variables and the links between them.
7. The method as claimed in claim 1 , wherein the technical entity is a motor vehicle.
8. A system for determining safety-critical output values (y 1 , y 2 , . . . , y n ) for a technical entity, said system comprising a data analysis device,
wherein the data analysis device is configured to (i) receive data and/or measured values for the entity and/or the surroundings of the entity, wherein the data/measured values describe at least one state and/or at least one feature of the entity and/or the surroundings of the entity and constitute input values (x), and (ii) process the input values (x) using a software application in order to determine at least one first output value (y 1 ) of the output values (y 1 , y 2 , . . . , y n ),
wherein the data analysis device is configured to process the input values (x) by:
(i) using a neural network having a plurality of layers (h θ (x)) with first learnable parameters (θ);
(ii) modifying a last layer or an additional layer of the neural network using a function (φ) so that the first output value (y 1 ) is located within a defined value range (C(s)) of at least one target parameter (s); and
(iii) determining the target parameter (s) and/or the value range (C(s)) of the target parameter (s) using further additional layers (k β (x)) of the neural network with second learnable parameters (β).
9. The system as claimed in claim 8 , wherein the target parameter (s) and/or the value range (C(s)) constitute at least one second output value (y 2 ) of the output values (y 1 , y 2 , . . . , y n ).
10. The system as claimed in claim 8 , wherein the data analysis device is configured to pass a representation (g(s)) formed from the target parameter (s) to the neural network as a further input value such that a calculation value (z) of the neural network also depends on the target parameter (s).
11. The system as claimed in claim 8 , wherein the data analysis device is configured to add a noise component (Δs) to the determined target parameter (s), the noise component (Δs) being a random number with the same dimension as the parameter (s).
12. The system as claimed in claim 9 , wherein the data analysis device is configured to separately or jointly use learnable parameters (γ) to determine the first output value (y 1 ) and the second output value (y 2 ).
13. The system as claimed in claim 8 , wherein use is made of a database that stores data relating to properties of the technical entity, images and characteristic variables and the links between them.
14. The system as claimed in claim 8 , wherein the technical entity is a motor vehicle.