Coprocessor Architecture for Mobile Platform Machine Vision

Resolve Bottlenecks,
Find Innovative Solutions
Generate Solutions

Solution Overview

Problem

Machine vision systems on mobile platforms face challenges with computationally expensive, power-intensive, and error-prone data processing, which are difficult to program and run in real-time, especially requiring deterministic timing for precise positioning.

Innovation Solution

An integrated system utilizing System on Chip (SOC) technologies to collect and process inertial, positional, and image data, distributing it to coprocessors for simplified and accurate local and global orientation, relieving the application processor of data acquisition workloads and reducing interrupts.

Engineering Contradictions & Design Principles

VSEngineering Contradiction Analysis

1Reliability

If machine vision systems use complicated components to calculate distance and object features based on relative movement through pixel arrays, then object avoidance and identification capabilities are improved, but computational complexity and processing time increase significantly

Engineering Contradiction:
Improveobject avoidance capabilityVSAvoidcomputational complexity
Core Design Contradiction:
ReliabilityVSDevice complexity

Solution Approach 1:

The system divides computational tasks between the application processor (running OS and high-level applications) and coprocessors (handling real-time sensor data acquisition and processing). This segmentation allows complex machine vision functions to be maintained while distributing computational burden to specialized hardware units that can operate in real-time without overwhelming the main processor.

Inventive Principle:
Principle #1Segmentation

Solution Approach 2:

Coprocessors act as intermediary hardware units between sensors and the application processor. They pre-process sensor data, perform initial object detection and tracking, and only transmit relevant results to the main processor, thereby reducing overall computational complexity while maintaining reliable object avoidance capabilities.

Inventive Principle:
Principle #24Intermediary (Mediator)

2Measurement precision

If machine vision systems perform computationally expensive data processing tasks, then processing accuracy is improved, but power consumption increases and real-time operation becomes difficult

Engineering Contradiction:
Improvepositioning accuracyVSAvoidpower consumption
Core Design Contradiction:
Measurement precisionVSUse of energy by moving object

Solution Approach 1:

Coprocessors are self-contained units that autonomously handle data acquisition and preliminary processing tasks without requiring continuous intervention from the main application processor. They maintain their own operational states, buffers, and processing pipelines, enabling them to service real-time processing needs independently and efficiently, thereby reducing overall system power consumption while maintaining positioning accuracy.

Inventive Principle:
Principle #25Self-service

Solution Approach 2:

The system replaces general-purpose software-based processing on the application processor with dedicated hardware coprocessors that perform specific processing functions. This substitution of specialized hardware for general-purpose computing reduces power consumption for equivalent processing tasks while maintaining or improving measurement precision through hardware-optimized algorithms.

Inventive Principle:
Principle #28Mechanics substitution (Replace mechanical system)

3Adaptability or versatility

If machine vision systems run multiple applications and control operations simultaneously, then system functionality is improved, but deterministic timing becomes difficult to achieve

Engineering Contradiction:
Improvesystem functionalityVSAvoidtiming determinism
Core Design Contradiction:
Adaptability or versatilityVSLoss of time

Solution Approach 1:

The system segments processing functions into real-time critical tasks (sensor data acquisition, immediate processing) handled by coprocessors, and non-critical tasks (high-level application logic, UI rendering) handled by the application processor running the OS. This segmentation isolates time-sensitive operations from OS scheduling variability, ensuring deterministic timing for critical functions while maintaining versatile system functionality through parallel operation of multiple applications.

Inventive Principle:
Principle #1Segmentation

Solution Approach 2:

Coprocessors serve as intermediary hardware that buffers and pre-processes sensor data before OS intervention is required. By handling data acquisition and initial processing independently of the OS schedule, coprocessors ensure that timing-critical operations maintain deterministic behavior even while the OS schedules multiple applications with varying priorities and timing requirements.

Inventive Principle:
Principle #24Intermediary (Mediator)

4Productivity

If mobile platforms move faster to improve productivity, then positioning precision requirements increase, but system errors increase due to linear errors from individual components

Engineering Contradiction:
Improvemobile platform speedVSAvoidpositioning precision
Core Design Contradiction:
ProductivityVSMeasurement precision

Solution Approach 1:

The system implements feedback loops where coprocessors continuously monitor sensor data and system state, dynamically adjusting processing parameters and error compensation algorithms in real-time. This feedback mechanism allows the system to maintain positioning precision at higher speeds by actively compensating for increased linear errors from individual components through real-time calibration and correction based on actual measured deviations.

Inventive Principle:
Principle #23Feedback

Data Source

PatentEP3098566B1System architecture for machine vision on moving platforms
Publication Date: 2020.12.16 GOODRICH CORP
  • EP3098566B1 patent drawingFigure 1
  • EP3098566B1 patent drawingFigure 2
  • EP3098566B1 patent drawingFigure 3

AI summary

A computer system for processing machine vision data and performing active control of a mobile platform includes a plurality of sensors configured to acquire inertial and positional data. The system further includes a first plurality of co-processors having a hardware logic configured to control the acquisition of the inertial data. A second plurality of sensors is configured to acquire image data related to the mobile platform. The system further includes a second plurality of co-processors having a hardware logic configured to control the acquisition of the image data. The system further includes state management logic to perform state management operations for the acquired inertial and positional data in a computer-readable memory. The state management is performed using a state vector. The state management logic coordinates sharing and updating the acquired machine vision data in a parallel fashion between the first and second co-processors.