Visual-Inertial Sensor Fusion for Positional Awareness
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Current technologies for providing fast, accurate, and reliable positional awareness to robots and wearable devices, such as VR/AR headsets, face challenges including high costs, power consumption, and limited accuracy, particularly in recognizing locations and obstructions quickly, with existing approaches like RFID/WiFi, depth sensors, and visual methods suffering from inefficiencies and scale ambiguity.
Innovation Solution
The implementation of a visual-inertial sensor system that combines grayscale and colored cameras with a multi-axis inertial measurement unit (IMU), offloading computational tasks to a low-power control unit, using stereo imaging capabilities, and employing techniques like image undistortion, feature detection, and sensor fusion to generate accurate and efficient positional data, reducing costs and power consumption.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Area of stationary object
If RFID/WiFi approaches are used for positional awareness, then coverage area is improved, but accuracy deteriorates and cost increases
Solution Approach 1:
The patent combines multiple sensing modalities (visual sensors, depth sensors, inertial sensors, RFID/WiFi receivers) into a unified positioning system. This sensor fusion approach allows the system to leverage the coverage advantages of RFID/WiFi while compensating for their accuracy limitations through complementary sensor data, particularly visual and depth information that provide precise spatial measurements.
Solution Approach 2:
The patent introduces visual features and depth map landmarks as intermediary reference points between the device and the environment. These visual intermediaries provide accurate positional feedback that mediates the less precise RFID/WiFi signals, enabling the system to achieve both wide coverage and high accuracy simultaneously by using visual landmarks as reference markers for positioning.
2Measurement precision
If depth sensor based approaches are used, then positioning accuracy is improved, but power consumption increases and cost increases
Solution Approach 1:
The patent implements periodic or event-triggered depth sensing rather than continuous operation. The depth sensor is activated at specific moments when visual features are detected or when positioning updates are needed, rather than operating continuously. This periodic activation maintains positioning accuracy when needed while dramatically reducing overall power consumption compared to continuous depth sensing.
Solution Approach 2:
The patent merges depth sensor data with visual sensor data and inertial sensor data in a unified positioning framework. By combining these complementary sensing modalities, the system can achieve accurate positioning using depth information only when necessary, while relying on lower-power visual and inertial sensors for continuous tracking, thus reducing overall power consumption while maintaining accuracy.
3Ease of manufacture
If visual approaches are used, then cost is reduced, but speed deteriorates leading to failure in fast motion applications
Solution Approach 1:
The patent performs preliminary processing of visual data by pre-detecting and tracking key features in the scene before full positioning computation is required. Visual features are identified and tracked in advance, creating a prepared data structure that enables faster real-time positioning updates during fast motion, thus maintaining both low cost and high speed performance.
Solution Approach 2:
The patent merges visual sensor data with high-speed inertial sensor data to compensate for the slower processing speed of visual approaches. The inertial sensors provide fast, high-speed motion information that complements the lower-speed but lower-cost visual data, enabling the system to achieve fast motion tracking capability without relying solely on expensive or fast-processing visual systems.
4Ease of manufacture
If visual approaches are used, then cost is reduced, but scale ambiguity deteriorates positioning accuracy
Solution Approach 1:
The patent combines visual sensor data with depth sensor data to resolve scale ambiguity. The depth sensor provides absolute distance measurements that establish the correct scale of the environment, while the visual sensor provides continuous tracking information. This fusion allows the low-cost visual system to achieve accurate scale information that would otherwise require expensive specialized sensors.
Solution Approach 2:
The patent uses depth map landmarks as intermediary reference objects that provide known geometric properties and scale information. These visual-depth hybrid intermediaries serve as reference markers with known dimensions, allowing the system to calibrate and resolve scale ambiguity in the visual data without requiring expensive specialized scale-reference equipment.
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.