Information processing device, in-vehicle control device, and vehicle control system

Patent metadata

Field	Value
Publication number	US-12020486-B2
Application number	US-201917413264-A
Country	US
Kind code	B2
Filing date	Dec 10, 2019
Priority date	Dec 12, 2018
Publication date	Jun 25, 2024
Grant date	Jun 25, 2024

Abstract

An information processing device includes a DNN operation unit that executes a DNN operation by a neural network including plural layers and a weight storage unit that stores a weight used in the DNN operation, and the DNN operation unit specifies data having a value larger than a predetermined threshold as operation target data among data input to a predetermined layer of the neural network, acquires a weight corresponding to the operation target data from the weight storage unit, and executes an operation of the predetermined layer based on the operation target data and the weight acquired from the weight storage unit.

First claim

The invention claimed is:

1. A vehicle control system comprising:
an in-vehicle control device comprising:
an information processing device comprising:
a DNN operation unit that executes a DNN operation by a neural network including plural layers; and
a weight storage unit that stores a weight used in the DNN operation,
wherein the DNN operation unit specifies data having a value larger than a predetermined threshold as operation target data among data input to a predetermined layer of the neural network, acquires a weight corresponding to the operation target data from the weight storage unit, and executes an operation of the predetermined layer based on the operation target data and the weight acquired from the weight storage unit;

an action plan formulation unit that formulates an action plan for a vehicle,
wherein the information processing device is mounted in the vehicle,
wherein the DNN operation unit executes the DNN operation based on sensor information related to the surrounding conditions of the vehicle, and
wherein the action plan formulation unit formulates the action plan for the vehicle based on the operation result of the DNN operation unit; and

a server device that can communicate with the in-vehicle control device,
wherein the server device determines the threshold, and
wherein the in-vehicle control device determines resource distribution for the DNN operation unit based on the threshold determined by the server device.

2. The vehicle control system of claim 1 , wherein the information processing device has a threshold of 0.

3. The vehicle control system of claim 1 , wherein the information processing device further comprises a threshold setting unit that sets the threshold,
wherein the DNN operation unit specifies data having a value larger than the threshold set by the threshold setting unit as the operation target data among the data input to the predetermined layer.

4. The vehicle control system of claim 3 , wherein for the information processing device,
the threshold setting unit sets the threshold for the current data based on an operation result of the DNN operation unit for the past data.

5. The vehicle control system of claim 1 , wherein the information processing device further comprises a data output limiting unit that, when the number of pieces of data in the operation result of the predetermined layer exceeds a predetermined limit setting value, limits the output of the operation result of the predetermined layer.

6. The vehicle control system of claim 5 , wherein for the information processing device,
the data output limiting unit selects data as an output limit target based on at least one of the operation order of each data, the absolute value of each data, and the magnitude of the weight corresponding to each data in the operation result of the predetermined layer.

7. The vehicle control system of claim 1 , wherein for the information processing device,
when the weight acquired from the weight storage unit is smaller than a predetermined weight threshold, the DNN operation unit omits the operation of the predetermined layer based on the weight.

Assignees

• Hitachi Astemo Ltd

Inventors

• Hirai Riu
• Ono Goichi
• Ito Hiroaki

Classifications

• G06N3/0464

  Convolutional networks [CNN, ConvNet] · CPC title

• G06N3/0495

  Quantised networks; Sparse networks; Compressed networks · CPC title

• G06N3/02

  Neural networks · CPC title

• G06N3/045

  Combinations of networks · CPC title

• G06N3/048

  Activation functions · CPC title