UAV Safe Landing Site Selection Using Depth Maps and Image Ranking
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Unmanned aerial vehicles (UAVs) face challenges in automatically identifying and navigating to a suitable landing location, especially when conventional navigation systems are compromised due to factors like inclement weather or electromagnetic interference.
Innovation Solution
A system that captures overlapping images of the landscape, generates a depth map to identify flat regions, and uses machine-learning-based models to rank landing zones for suitability, enabling the UAV to autonomously navigate and land on the most suitable location.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Reliability
If conventional navigation systems (GPS, visual-based) are used for UAV landing, then navigation accuracy is maintained under normal conditions, but the system becomes vulnerable to failure under adverse conditions (inclement weather, electromagnetic interference)
Solution Approach 1:
The patent introduces an intermediary vision-based navigation system that mediates between the failed conventional GPS/visual navigation systems and the landing execution. This intermediary system uses onboard cameras and image processing to independently determine position and navigate to landing zones, bridging the gap when primary systems fail under adverse conditions.
Solution Approach 2:
The system changes the operational parameters of navigation by switching from satellite-based GPS signals to visual-based image processing parameters. When conventional systems fail, the UAV transitions to using computer vision parameters (image features, depth estimation, optical flow) that are immune to electromagnetic interference and weather conditions affecting radio signals.
2Extent of automation
If automatic landing systems are implemented, then operational autonomy is improved, but the complexity of the navigation and landing system increases
Solution Approach 1:
The UAV performs self-service navigation and landing by autonomously identifying suitable landing zones, calculating navigation paths, and executing landing maneuvers without external intervention. The system uses its own onboard sensors and processing capabilities to independently complete the entire landing sequence, reducing the need for complex external control infrastructure.
Solution Approach 2:
The autonomous landing system is segmented into distinct functional modules: image capture module, landing zone identification module (using machine learning models), navigation path planning module, and landing execution module. This segmentation allows each component to be optimized independently and simplifies the overall system architecture by dividing complex autonomous landing into manageable functional segments.
3Reliability
If machine learning models are used to identify suitable landing zones, then landing safety is improved, but computational requirements and processing time increase
Solution Approach 1:
The system performs preliminary identification of potential landing zones during the approach phase using machine learning models trained on landscape imagery. By pre-identifying and ranking suitable landing zones before final descent, the system reduces real-time processing requirements during critical landing moments, balancing safety with timing constraints.
Applied Scientific Principles
This section explains which scientific principles are used to turn an abstract innovation direction into a practical engineering solution.
Function Achieved in This Case
Enables UAVs to automatically and safely land even when primary navigation systems fail, by identifying suitable flat regions and ranking them for landing quality, ensuring reliable operation in various conditions.
Implementation Method 1
generate a depth map based on a change in relative position among corresponding (e.g., matching) pixels between the images
Data Source
Figure 1
Figure 2
Figure 3
AI summary
An unmanned aerial vehicle (UAV) navigation system is configured to automatically identify a suitable UAV landing site, for example, in forced-landing (e.g., emergency) scenarios. In some examples, the system receives two or more overlapping images depicting a landscape underneath an airborne unmanned aerial vehicle (UAV); generates, based on the two or more overlapping images, a depth map for the landscape; identifies, based on the depth map, regions of the landscape having a depth variance below a threshold value; determines, for each of the regions of the landscape having a depth variance below the threshold value, a landing zone quality score indicative of the depth variance and a semantic type of the region of the landscape; identifies, based on the landing zone quality scores, a suitable location for landing the UAV; and causes the UAV to automatically navigate toward and land on the suitable location.