Rotorcraft Image-Based Proximity Sensing for Rotor Strike Avoidance
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Piloting a rotorcraft is challenging due to the difficulty in detecting low-visibility objects, such as fences, power lines, and rotor blades of other aircraft, which increases the risk of rotor blade strikes.
Innovation Solution
The use of artificial intelligence (AI) models to process image information from sensors mounted on the rotorcraft, detecting objects and determining their distance to facilitate avoiding collisions with physical objects in the surrounding environment.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Difficulty of detecting and measuring
If image sensors and AI models are used to detect objects and determine distance, then the detection capability for low-visibility objects is improved, but the device complexity increases
Solution Approach 1:
The patent introduces image sensors as intermediary devices that capture visual information of the surrounding environment, and AI models as computational intermediaries that process this information to detect objects and estimate distances. These intermediaries bridge the gap between the rotorcraft's need for environmental awareness and the physical limitations of human pilots and traditional sensors, enabling detection of low-visibility objects without requiring direct human observation or complex mechanical sensing systems
Solution Approach 2:
The patent replaces mechanical and human-based detection methods with electronic and computational systems. Instead of relying on human visual detection or traditional mechanical sensors, the system uses image sensors coupled with AI-based object detection and depth estimation algorithms. This substitution transitions from mechanical/optical direct detection to computational image analysis, reducing the need for complex mechanical sensing apparatus while improving detection capability
2Reliability
If multiple image sensors and AI processing are implemented, then the safety and collision avoidance capability are improved, but the use of energy increases
Solution Approach 1:
The patent implements selective processing where the AI system focuses computational resources on analyzing regions of interest within the captured images. Rather than processing every pixel uniformly, the system identifies potential objects and concentrates processing power on those areas, using partial action to achieve sufficient detection accuracy while reducing overall computational energy consumption. The depth estimation is also performed selectively based on detected object significance
Solution Approach 2:
The system uses the image sensors to serve multiple functions simultaneously - both object detection and depth estimation are derived from the same image data without requiring separate sensing systems. The AI models process the image information to generate both object identification and proximity information, allowing the system to serve its safety function efficiently by maximizing the utility of each sensor and computation unit
Data Source
Figure 1
Figure 2A~2B
Figure 2C
AI summary
In certain embodiments, a method includes accessing image information generated by one or more image sensors configured to generate the image information for a surrounding environment of a rotorcraft. The method includes causing one or more Al models to process the image information to generate proximity information for the image information. The proximity information includes depth measurements for one or more image objects from the image information that correspond to one or more physical objects in the surrounding environment of the rotorcraft. The method includes initiating, in response to generating the proximity information from the image information, further analysis of the proximity information for the one or more image objects to facilitate avoiding, based on the proximity information, a collision of a rotor blade of the rotorcraft with the one or more physical objects that correspond to the one or more image objects.