Heterogeneous Compute Platform Safety Through Split Algorithm Execution

Resolve Bottlenecks,
Find Innovative SolutionsGenerate Solutions

Solution Overview

Problem

Autonomous and semi-autonomous vehicles face challenges in achieving algorithm safety on heterogeneous compute platforms, where non-safety compliant computing units are often used for efficiency but lack compliance with automotive safety integrity levels, posing risks in safety-critical functions.

Innovation Solution

The method involves determining whether a non-safety compliant computing unit is preferred for executing an algorithm, modifying its execution to leverage both non-safety and safety compliant units, creating lighter versions of the algorithm, and generating outputs on both units to ensure safety compliance, with mechanisms to verify output matching and raise alarms for discrepancies.

Engineering Contradictions & Design Principles

VSEngineering Contradiction Analysis

1Productivity

If non-safety compliant computing units are used to execute algorithms, then processing efficiency is improved, but algorithm safety and compliance with automotive safety integrity levels deteriorate

Engineering Contradiction:

Improveprocessing efficiencyVSAvoidalgorithm safety

Core Design Contradiction:

ProductivityVSReliability

Solution Approach 1:

The algorithm is divided into multiple portions, with critical safety-related portions executed on safety-compliant computing units and non-critical portions executed on non-safety compliant units. This segmentation allows the system to leverage the high processing efficiency of non-safety compliant units while ensuring that safety-critical functions maintain compliance with automotive safety integrity levels.

Inventive Principle:

Principle #1Segmentation

Solution Approach 2:

Different computing units with different safety compliance levels are assigned to different portions of the algorithm based on their safety characteristics and performance capabilities. Safety-compliant units handle portions requiring guaranteed safety, while non-safety compliant units handle portions where performance is prioritized and safety risks are lower.

Inventive Principle:

Principle #3Local quality

2Reliability

If algorithm execution is modified to use both non-safety and safety compliant computing units, then system reliability is improved, but device complexity increases

Engineering Contradiction:

Improvesystem reliabilityVSAvoidexecution management complexity

Core Design Contradiction:

ReliabilityVSDevice complexity

Solution Approach 1:

The system performs preliminary analysis to identify which portions of the algorithm are safety-critical and require execution on safety-compliant units. This upfront classification simplifies the overall execution management by establishing clear guidelines for portion allocation before runtime execution begins.

Inventive Principle:

Principle #10Preliminary action

Solution Approach 2:

A portion identification module acts as an intermediary that analyzes the algorithm, identifies safety-critical portions, and directs them to appropriate computing units. This intermediary component manages the complexity of coordinating multiple computing units with different safety compliance levels.

Inventive Principle:

Principle #24Intermediary (Mediator)

Data Source

PatentUS11834071B2System to achieve algorithm safety in heterogeneous compute platform

Publication Date: 2023.12.05 QUALCOMM INC

AI summary

Methods, system, non-transitory media, and devices for supporting safety compliant computing in a heterogeneous computing system, such as a vehicle heterogeneous computing system are disclosed. Various aspects include methods enabling a vehicle, such as an autonomous vehicle, a semi-autonomous vehicle, etc., to achieve algorithm safety for various algorithms on a heterogeneous compute platform with various safety levels.