Compute Task Distribution for Autonomous Vehicle Hardware

Resolve Bottlenecks,
Find Innovative SolutionsGenerate Solutions

Solution Overview

Problem

Existing computing systems face challenges in optimizing compute tasks for execution on specific hardware topologies, particularly in autonomous vehicle computing systems, where achieving robust safety integrity levels is crucial.

Innovation Solution

A computing system that determines a set of weighted parameters for a given hardware topology, optimally distributes runnables of a compute graph, and optimizes data positioning in memory components to facilitate efficient execution, using techniques such as the traveling salesman algorithm and learning-based approaches.

Engineering Contradictions & Design Principles

VSEngineering Contradiction Analysis

1Loss of time

If compute tasks are distributed on a hardware topology without optimization, then device complexity is reduced, but latency and power consumption increase

Engineering Contradiction:

ImprovelatencyVSAvoiddistribution optimization complexity

Core Design Contradiction:

Loss of timeVSDevice complexity

Solution Approach 1:

The system changes parameters by determining weighted parameters for the hardware topology that include latency, bandwidth, memory, power usage, computing power, compute units, hardware age, hardware wearing, and thermal cooling. These parameter changes enable optimized distribution of runnables across the hardware topology to minimize latency while accounting for multiple performance factors.

Inventive Principle:

Principle #35Parameter changes

Solution Approach 2:

The patent replaces manual or heuristic task distribution methods with an automated optimization system that uses algorithms to determine optimal runnable distribution. This substitution of the distribution mechanism enables systematic optimization of latency and power consumption across the hardware topology.

Inventive Principle:

Principle #28Mechanics substitution (Replace mechanical system)

2Use of energy by moving object

If compute tasks are distributed to minimize power consumption, then energy efficiency improves, but execution time may increase

Engineering Contradiction:

Improvepower consumptionVSAvoidexecution speed

Core Design Contradiction:

Use of energy by moving objectVSProductivity

Solution Approach 1:

The optimization system incorporates power usage as a weighted parameter alongside execution time considerations. By adjusting the weights of different parameters including power consumption and latency, the system can balance energy efficiency with execution speed based on specific operational requirements and hardware state.

Inventive Principle:

Principle #35Parameter changes

Solution Approach 2:

The system dynamically determines optimal runnable distribution based on current hardware topology parameters and runnable requirements. This dynamic optimization allows the system to adapt to changing conditions, balancing power consumption and execution speed in real-time rather than using static allocation.

Inventive Principle:

Principle #15Dynamics

3Productivity

If data is positioned in memory components for optimal execution, then computing efficiency improves, but memory management complexity increases

Engineering Contradiction:

Improvecomputing efficiencyVSAvoidmemory management complexity

Core Design Contradiction:

ProductivityVSDevice complexity

Solution Approach 1:

The system performs preliminary action by optimizing data positioning in memory components before execute the runnables. By determining optimal data positions in advance based on the weighted parameters and runnable requirements, the system improves computing efficiency while the complexity is managed through automated optimization rather than manual memory management.

Inventive Principle:

Principle #10Preliminary action

4Adaptability or versatility

If the system reevaluates optimal distribution frequently, then adaptability to hardware changes improves, but computational overhead increases

Engineering Contradiction:

Improveadaptability to hardware changesVSAvoidcomputational overhead

Core Design Contradiction:

Adaptability or versatilityVSUse of energy by stationary object

Solution Approach 1:

The system uses dynamic optimization by reevaluating the optimal distribution of runnables based on updated weighted parameters for the hardware topology. This allows the system to adapt to hardware changes such as aging, wearing, or thermal conditions while managing computational overhead through algorithmic efficiency.

Inventive Principle:

Principle #15Dynamics

Solution Approach 2:

The optimization process incorporates feedback by determining updated weighted parameters and reevaluating the optimal distribution based on current hardware state and runnable requirements. This feedback mechanism enables the system to adapt to changes while the computational overhead is justified by the improved performance and reliability.

Inventive Principle:

Principle #23Feedback

Data Source

PatentUS20250045113A1Systems and methods of optimizing compute tasks

Publication Date: 2025.02.06 MERCEDES BENZ GROUP AG

AI summary

A computing system can execute a heuristic technique (e.g., traveling salesman algorithm) and/or a learning-based technique to determine an optimal distribution of compute tasks for execution on a given hardware topology.