Multi-Sensor UAV Positioning for GPS-Denied Indoor Navigation
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Conventional unmanned aerial vehicles (UAVs) struggle to accurately recognize their position indoors without GPS data, as existing methods are costly, require manual flight, and are inefficient with increasing target area size, and fail to determine absolute or relative positions when edge or line features are not extractable.
Innovation Solution
An UAV equipped with an inertia sensor, tag recognition sensor, and image sensor, using an extended Kalman filter to estimate position and control movement, which includes acquiring inertia information, recognizing tags for absolute position, and applying visual odometry data to the filter for robust position recognition.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If a 3D lidar is used to generate a 3D map for position recognition, then position recognition capability is improved, but cost and device complexity increase significantly
Solution Approach 1:
The patent combines multiple sensors (inertial sensor, tag recognition sensor, image sensor) into an integrated sensor system that works together through data fusion. This merging approach achieves accurate position recognition without requiring expensive 3D lidar, as the combined data from cheaper individual sensors produces comparable results through the extended Kalman filter algorithm.
Solution Approach 2:
The sensor system performs multiple functions simultaneously: the inertial sensor tracks motion, the tag recognition sensor provides absolute position references, and the image sensor captures visual features. This multi-functional system replaces the single-function 3D lidar while achieving the same position recognition goal through diverse sensor inputs.
2Measurement precision
If a 3D map is generated for position recognition, then position accuracy is improved, but memory and calculation time increase rapidly as target area becomes larger
Solution Approach 1:
Instead of generating and processing complete 3D maps of the entire target area, the system extracts only the necessary position information through local feature matching and tag recognition. The extended Kalman filter processes incremental sensor data rather than entire map datasets, significantly reducing memory requirements and calculation time while maintaining position accuracy.
Solution Approach 2:
The system uses partial action by processing only the minimal sensor data needed for position estimation at each time step rather than processing complete environmental maps. This approach processes a subset of available information (current sensor readings) sufficient for the task, avoiding the computational burden of processing all possible data.
3Loss of information
If manual flight is used to generate position data, then position information can be obtained, but ease of operation deteriorates due to requiring manual intervention
Solution Approach 1:
The UAV performs self-positioning autonomously using its onboard sensor system. The inertial sensor continuously tracks motion, the tag recognition sensor automatically identifies position references, and the extended Kalman filter autonomously fuses data to estimate position. This self-service capability eliminates the need for manual flight operations to generate position data, improving ease of operation while maintaining accurate position information.
4Measurement precision
If edge or line features are used for position recognition, then relative position can be determined, but reliability decreases when edges or lines cannot be extracted
Solution Approach 1:
The system introduces tags as intermediary position references that are deliberately placed in the environment. These tags serve as reliable mediators between the UAV and the environment, providing unambiguous absolute position information that doesn't depend on extracting edges or lines from natural features. The tag recognition sensor reliably detects these intermediaries even in featureless environments.
Solution Approach 2:
The system changes the parameter being measured from extracting geometric features (edges, lines) to recognizing predefined tag patterns. This parameter change from feature extraction to pattern recognition improves reliability because tags provide consistent, high-contrast visual targets that are easier to reliably detect than natural environmental features, especially in controlled indoor environments.
Data Source
AI summary
An unmanned aerial vehicle may include: a sensor part configured to acquire inertia information or position information of the unmanned aerial vehicle; and a controller. The controller is configured to estimate the position of the unmanned aerial vehicle by applying the information acquired by the sensor part to an extended Kalman filter and control movement of the unmanned aerial vehicle, based on the estimated position of the unmanned aerial vehicle. The sensor part includes: an inertia sensor configured to acquire the inertia information of the unmanned aerial vehicle; a tag recognition sensor configured to recognize a tag attached to a rack and acquire absolute position information of the unmanned aerial vehicle; and an image sensor attached to the unmanned aerial vehicle so as to acquire an image of the movement environment of the unmanned aerial vehicle.