Stereo Obstacle Avoidance for Autonomous Aircraft in Unstructured Flight
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Conventional perception algorithms for autonomous vehicles, particularly UAVs, face challenges in detecting obstacles in unstructured environments and degraded conditions, such as weather or radiation exposure, and struggle with three-dimensional obstacle avoidance and handling small objects, leading to accumulated errors and high performance costs.
Innovation Solution
An obstacle detection and avoidance system (ODOA) that uses image sensing to generate a potential field correction value, allowing autonomous vehicles to adjust their travel path and avoid collisions by processing disparity images and occupancy maps, without requiring external communication, and is suitable for vehicles with hardware constraints.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Ease of operation
If conventional perception algorithms map the environment and maintain a state of the world, then navigation capability is improved, but accumulated error increases and performance costs rise
Solution Approach 1:
The patent segments the environment into discrete depth bins and occupancy grids rather than maintaining a continuous world model. This segmentation allows the system to process environmental data in manageable discrete units, reducing accumulated error while maintaining navigation capability through localized spatial reasoning.
Solution Approach 2:
The patent introduces depth maps and occupancy grids as intermediary representations between the sensor data and navigation decisions. These intermediaries transform complex environmental information into simplified spatial probability distributions, reducing computational performance costs while maintaining reliable navigation through probabilistic reasoning.
2Productivity
If the minimum grid scale of the map is increased to reduce performance requirements, then processing speed is improved, but ability to handle small objects deteriorates
Solution Approach 1:
The patent applies local quality by using variable-resolution occupancy grids where different spatial regions can have different grid resolutions. This allows finer grid scales in regions with small objects or high interest while using coarser scales in less critical areas, maintaining small object detection capability without sacrificing overall processing speed.
Solution Approach 2:
The patent transitions from two-dimensional map representations to three-dimensional occupancy grids with depth information. This additional dimension allows the system to detect small objects more effectively by providing depth context, while the discrete binning approach maintains processing efficiency through reduced computational complexity.
3Ease of operation
If perception algorithms are designed for structured roads and lanes, then automotive navigation is improved, but adaptability to unstructured environments deteriorates
Solution Approach 1:
The patent implements universal perception algorithms based on occupancy grids and depth mapping that can function across multiple vehicle types and environments. The same core algorithms handle both structured road navigation and unstructured terrain exploration, achieving multi-functionality without requiring environment-specific customization.
Solution Approach 2:
The patent uses parameter changes by adjusting occupancy grid resolution, depth binning intervals, and sensor fusion weights based on environmental conditions and vehicle type. This allows the system to adapt its perception parameters dynamically, maintaining optimal performance whether navigating structured roads or unstructured terrain.
4Stability of the object's composition
If conventional algorithms are used in degraded environments such as weather or radiation exposure, then standard operation is maintained, but detection capability deteriorates
Solution Approach 1:
The patent implements feedback mechanisms where the occupancy grid system continuously updates its environmental model based on sensor measurements and vehicle state. This feedback loop allows the system to compensate for degraded detection conditions by using historical occupancy information and probabilistic reasoning to maintain detection capability in adverse weather or radiation environments.
Solution Approach 2:
The patent applies beforehand cushioning by maintaining predictive occupancy models that anticipate obstacles based on vehicle trajectory and environmental patterns. This predictive capability provides a buffer against detection failures in degraded conditions, allowing the system to prepare avoidance maneuvers before obstacles are clearly detected.
Data Source
AI summary
A method of providing a collision avoiding travel path for an autonomous vehicle. A sensor system obtains stereo image data of a scene in the environment ahead of the normal travel path. This image data is used to generate a disparity image. The disparity image is processed to generate an occupancy map that assigns values to areas of the scene based on levels of visual clutter. The occupancy map is then converted to a potential field, which assigns each pixel in the scene with a force value that corresponds to its proximity to one or more obstacles. These force value are summed and used to modify the vehicle's path is a collision is likely.