Visual-Inertial Mapping for Fast Low-Power Positional Awareness
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Current technologies face challenges in providing fast, accurate, and reliable positional awareness for autonomous robots and wearable devices, such as virtual reality headsets, due to limitations in existing sensing methods like RFID/WiFi, depth sensors, and visual approaches, which are often expensive, power-intensive, or suffer from scale ambiguity and speed limitations.
Innovation Solution
The implementation of a visual-inertial sensor system that combines stereo imaging sensors with low-end cameras and a low-power control unit to process image and inertial data efficiently, offloading computational tasks from the main processor and using a hybrid point grid for localization and mapping, enabling cost-effective and energy-efficient positional awareness.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If depth sensor based approaches are used, then positional awareness accuracy is improved, but cost and power consumption increase
Solution Approach 1:
The patent combines visual sensors (cameras) with inertial sensors (accelerometers, gyroscopes) into a unified visual-inertial system. This fusion allows the system to achieve depth sensor-level positional awareness accuracy without the high power consumption and cost, by merging complementary sensing modalities that together provide robust 6-DOF pose estimation
Solution Approach 2:
The patent replaces expensive depth sensors (which use active mechanical or optical systems like LIDAR or structured light) with a passive visual-inertial system. The inertial sensors provide direct motion measurements that substitute for the complex depth sensing mechanism, while visual features provide environmental context, achieving the same function with lower cost and power
2Ease of manufacture
If visual approaches are used, then cost is reduced, but speed and reliability deteriorate in fast motion applications
Solution Approach 1:
The patent performs preliminary action by using inertial sensors to predict device motion and pre-compute expected feature positions before visual processing. This allows the visual system to operate at lower speeds while maintaining accuracy, as the inertial data provides advance information about scene changes, enabling cost-effective visual sensors to keep pace with fast motion
Solution Approach 2:
The patent implements feedback by continuously fusing visual measurements with inertial predictions, where inertial data provides high-speed feedback about device motion that compensates for the slower visual processing rate. This feedback loop maintains reliability in fast motion applications while allowing the use of lower-cost visual sensors
3Ease of manufacture
If visual approaches are used, then cost is reduced, but scale ambiguity increases
Solution Approach 1:
The patent merges visual features with inertial measurements to resolve scale ambiguity. While visual monocular approaches suffer from scale ambiguity, the inertial sensors provide direct measurements of acceleration and angular velocity that are scale-invariant. By combining these modalities, the system recovers metric scale information without increasing cost
4Measurement precision
If traditional SLAM techniques are used, then positional awareness capability is improved, but speed and reliability deteriorate
Solution Approach 1:
The patent extracts and utilizes high-frequency inertial measurements from the sensor system, separating this fast data stream from the lower-frequency visual processing pipeline. By extracting inertial information for immediate pose updates, the system achieves both the positional awareness capability of traditional SLAM and the processing speed required for real-time autonomous operation
Data Source
AI summary
The described positional awareness techniques employing visual-inertial sensory data gathering and analysis hardware with reference to specific example implementations implement improvements in the use of sensors, techniques and hardware design that can enable specific embodiments to provide positional awareness to machines with improved speed and accuracy.