UAV Sensor Data Set Generation for Moving-Object SLAM Evaluation
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing methods for generating data sets for evaluating UAV position estimation algorithms, such as SLAM and VIO, are inefficient and inaccurate, particularly in simulating environments with moving objects, leading to difficulties in comparing algorithm performances and matching sensor measurement periodicities.
Innovation Solution
A method and apparatus for generating data sets by obtaining flight environment information, including moving objects, and simulating sensor data with varying periodicities to create sets with and without moving objects, allowing for accurate analysis of algorithm robustness and performance.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Productivity
If data sets are obtained using a virtual environment, then data acquisition speed and efficiency are improved, but accuracy of UAV position measurement deteriorates
Solution Approach 1:
The patent creates a virtual copy of the real UAV flight environment, including virtual UAVs, sensors, and moving objects. This virtual environment replicates real-world conditions while allowing efficient data generation. The virtual sensors in the simulation copy real sensor behaviors and measurement periodicities, enabling both speed and accuracy.
Solution Approach 2:
The system dynamically adjusts parameters such as sensor measurement periodicities, moving object velocities, and environmental conditions to match real-world scenarios. By carefully controlling these parameters in the virtual environment, the patent maintains measurement accuracy while benefiting from fast virtual data generation.
2Adaptability or versatility
If data sets are generated by users through manual processes, then flexibility in controlling surrounding situations is improved, but process complexity and time consumption increase
Solution Approach 1:
The patent implements a dynamic virtual environment where moving objects can be programmatically controlled to exhibit various behaviors (random motion, periodic motion, tracking). This allows flexible scenario creation through code rather than manual intervention, reducing complexity while maintaining adaptability.
Solution Approach 2:
The system automatically generates data sets by executing simulation programs that self-manage the virtual environment, sensor data collection, and data set compilation. This automation eliminates manual complexity while preserving the ability to control surrounding situations through programmatic interfaces.
3Measurement precision
If data sets include various information for accurate algorithm evaluation, then evaluation accuracy is improved, but data set complexity increases
Solution Approach 1:
The patent segments the data set into distinct components: flight data (UAV position, attitude), sensor data (camera, lidar, IMU with different periodicities), and environment data (moving objects, static objects). This structured segmentation organizes complex information into manageable sections while preserving all necessary evaluation details.
Solution Approach 2:
The patent creates a universal data set structure that can evaluate multiple algorithms (SLAM, VIO, object detection) simultaneously. The same data set format accommodates different sensor types and algorithm requirements, reducing overall complexity through standardization while maintaining comprehensive evaluation capability.
Data Source
AI summary
A method for generating data sets of a UAV, performed by a data set generation apparatus, may comprise obtaining flight environment information including a moving object; obtaining flight data including first physical information and first sensor information of the UAV in a flight environment based on the flight environment information; obtaining second physical information and second sensor information of the UAV when the moving object exists, based on the flight data, and generating a first data set based on the second physical information and the second sensor information; obtaining third physical information and third sensor information of the UAV when the moving object does not exist, based on the flight data, and generating a second data set based on the third physical information and the third sensor information; and combining the first data set and the second data set to generate a third data set.