Local Reconstruction for Low-Latency Remote Rendering

Resolve Bottlenecks,
Find Innovative SolutionsGenerate Solutions

Solution Overview

Problem

Conventional remote rendering techniques face inefficiencies due to latency, bandwidth limitations, and reduced quality of experience in interactive digital content delivery, particularly in immersive environments, due to encoding and decoding processes that compromise high dynamic range support and geometric shading capabilities.

Innovation Solution

Implementing local reconstruction techniques at the client device using machine-learning models to reconstruct digital content, including high dynamic range pixels, geometry buffers, and illumination effects, leveraging generative artificial intelligence to enhance interactivity and quality of experience.

Engineering Contradictions & Design Principles

VSEngineering Contradiction Analysis

1Use of energy by moving object

If remote rendering is used to offload rendering computations to a producer device, then computational resources are optimized, but latency increases and interactivity deteriorates

Engineering Contradiction:

Improvecomputational resourcesVSAvoidlatency

Core Design Contradiction:

Use of energy by moving objectVSLoss of time

Solution Approach 1:

The rendering process is segmented into two parts: heavy rendering computations are performed remotely at the producer device, while lightweight local reconstruction is performed at the client device using machine learning models. This segmentation allows each device to perform tasks suited to its capabilities, reducing overall latency while optimizing computational resource usage.

Inventive Principle:

Principle #1Segmentation

Solution Approach 2:

A machine learning-based local reconstruction model acts as an intermediary between the remotely rendered digital content and the final display. This intermediary component reconstructs high dynamic range pixels, geometry buffers, and illumination effects locally, eliminating the need for high-bandwidth transmission of complete render data and reducing latency.

Inventive Principle:

Principle #24Intermediary (Mediator)

2Loss of energy

If data compression is used to reduce bandwidth requirements, then transmission efficiency improves, but quality of digital content deteriorates

Engineering Contradiction:

ImprovebandwidthVSAvoidquality of digital content

Core Design Contradiction:

Loss of energyVSManufacturing precision

Solution Approach 1:

Instead of transmitting the complete high-quality rendered content, the system transmits a compressed representation and uses local machine learning models to generate a high-quality copy of the essential visual information. This copying approach maintains quality while significantly reducing bandwidth requirements.

Inventive Principle:

Principle #26Copying

Solution Approach 2:

The system changes the parameter representation by transforming detailed rendered pixel data into a compressed format that can be efficiently transmitted, then locally reconstructs the high-quality visual parameters (high dynamic range pixels, geometry buffers, illumination effects) using machine learning models, achieving both bandwidth efficiency and content quality.

Inventive Principle:

Principle #35Parameter changes

3Adaptability or versatility

If encoding and decoding processes are used to transmit digital content, then content delivery is enabled, but processing time increases and interactivity decreases

Engineering Contradiction:

Improvecontent deliveryVSAvoidinteractivity

Core Design Contradiction:

Adaptability or versatilityVSProductivity

Solution Approach 1:

The system extracts only the essential rendered content data from the complete rendering pipeline and transmits this minimized representation to the client device. The local machine learning model then reconstructs the remaining visual information, significantly reducing processing time and enabling real-time interactivity while maintaining content delivery capabilities.

Inventive Principle:

Principle #2Taking out (Extraction)

Solution Approach 2:

The producer device performs preliminary rendering computations and prepares compressed representations of the content in advance. The client device then uses local machine learning models to rapidly reconstruct the content without waiting for complete encoding and decoding cycles, reducing processing time and improving interactivity.

Inventive Principle:

Principle #10Preliminary action

Data Source

PatentUS20250308079A1Local Reconstruction of Remotely Rendered Digital Content

Publication Date: 2025.10.02 ADVANCED MICRO DEVICES INC

AI summary

Local reconstruction techniques of remotely rendered digital content are described. In one or more examples, a device includes a decoder implemented in hardware and configured to generate a decoded digital image from an encoded digital image and a renderer implemented in hardware and configured to reconstruct a digital image from the decoded digital image by rendering the decoded digital image using a machine-learning model.