Visual-Inertial Navigation With EKF for GPS-Denied Localization
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing navigation technologies, such as GPS-based systems, face challenges in urban and indoor areas due to poor signal reception and inability to provide close-quarter navigation for accident avoidance. Additionally, they suffer from sensor errors and inability to detect humans or obstacles.
Innovation Solution
A vision-aided inertial navigation system that processes visual information from a camera and inertial sensor data using an Extended Kalman filter (EKF) to estimate pose and localization information, achieving linear computational complexity with the number of tracked features.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Adaptability or versatility
If GPS-based navigation is used, then navigation coverage is provided, but signal reception deteriorates in urban and indoor areas
Solution Approach 1:
The patent combines GPS reception with visual sensor data and inertial sensor data to form a hybrid navigation system. The visual-inertial odometry module processes camera images and inertial measurements independently, then integrates results with GPS data through a Kalman filter, enabling continuous navigation even when GPS signals are blocked in urban canyons or indoor environments.
Solution Approach 2:
The visual-inertial odometry system acts as an intermediary when GPS signals are unavailable. The system detects feature points in visual data, tracks their motion, and computes pose estimates that serve as a substitute for GPS positioning, allowing seamless transition between GPS-based and vision-based navigation modes.
2Measurement precision
If the number of tracked features increases, then localization precision improves, but computational complexity increases
Solution Approach 1:
The patent segments the computational task by separating feature detection, feature tracking, and pose estimation into distinct modules. The visual-inertial odometry algorithm processes visual data and inertial data through separate measurement models, then combines them through Kalman filtering. This modular approach allows the system to handle numerous feature points efficiently by processing them through standardized computational pipelines.
Solution Approach 2:
The system dynamically adjusts the number of tracked feature points based on computational resources and environmental conditions. The Kalman filter uses adaptive state vectors that can accommodate varying numbers of features, and the measurement noise covariance matrices are adjusted based on feature quality and tracking reliability, optimizing the balance between precision and computational load.
3Adaptability or versatility
If inertial sensors are used for close-quarter navigation, then navigation capability improves, but sensor errors accumulate
Solution Approach 1:
The system implements feedback through the Kalman filter, which continuously compares inertial sensor predictions with actual visual measurements. The filter uses the difference between predicted and observed feature point positions to correct inertial drift, providing real-time feedback that prevents error accumulation during close-quarter maneuvers where inertial data is critical.
Solution Approach 2:
The system prepares for inertial error accumulation by pre-integrating visual constraints into the navigation solution. The visual-inertial odometry algorithm establishes geometric constraints from tracked feature points before they can cause significant drift, cushioning against error accumulation by continuously anchoring the inertial solution to visually-verified positions.
Data Source
AI summary
Localization and navigation systems and techniques are described. An electronic device comprises a processor configured to maintain a state vector storing estimates for a position of the electronic device at poses along a trajectory within an environment along with estimates for positions for one or more features within the environment. The processor computes, from the image data, one or more constraints based on features observed from multiple poses of the electronic device along the trajectory, and computes updated state estimates for the position of the electronic device in accordance with the motion data and the one or more computed constraints without computing updated state estimates for the features for which the one or more constraints were computed.