Autonomous Drone Inspection Navigation With Real-Time Pose Refinement
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Autonomous vehicles, particularly drones, face challenges in navigating and performing tasks in real-world environments with unknown or changing conditions, requiring improved control and navigation systems to enhance automation and reduce operator intervention.
Innovation Solution
The development of an autonomous vehicle system that includes a navigation system capable of generating and combining in-flight models using feature points and lines, with a CNN architecture for pose optimization and localization, enabling the vehicle to adapt and refine its flight plan in real-time based on sensor data and environmental changes.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Productivity
If the degree of automation of drones is increased to perform inspection tasks without operator intervention, then productivity and task repeatability are improved, but the system becomes more complex and difficult to control in unknown environments
Solution Approach 1:
The navigation system is divided into multiple independent modules: a navigation model generation module that creates 2D navigation models from sensor data, a pose optimization module that refines vehicle positioning, and a mapping module that builds environmental maps. Each module operates independently but contributes to the overall autonomous navigation function, allowing the complex system to be managed through modular components.
Solution Approach 2:
The system performs preliminary actions by generating navigation models and environmental maps before the inspection task begins. The navigation model generation module creates initial 2D representations of the environment, and the mapping module pre-processes spatial information, enabling the autonomous vehicle to navigate without real-time operator intervention.
2Adaptability or versatility
If the drone navigates in real-world environments with unknown or changing conditions, then adaptability is improved, but measurement precision and localization accuracy deteriorate
Solution Approach 1:
The pose optimization module implements feedback by continuously comparing the autonomous vehicle's estimated position with observed feature points in the environment. The system refines localization accuracy by iteratively optimizing pose estimates based on sensor measurements and navigation models, allowing accurate navigation even in changing environmental conditions.
Solution Approach 2:
The navigation system dynamically adapts to environmental changes by continuously updating the 2D navigation models and re-optimizing pose estimates in real-time. The system transitions from static pre-planned paths to dynamic route adjustments based on current sensor data and environmental conditions, maintaining localization accuracy despite environmental variability.
3Device complexity
If model-based camera pose estimation is used for autonomous navigation, then device complexity is reduced, but measurement precision deteriorates when feature points are out-of-view
Solution Approach 1:
The system addresses the out-of-view feature point problem by transitioning from 2D image plane coordinates to 3D spatial coordinates. The navigation model generation module creates 2D navigation models that are then transformed into 3D navigation models, allowing the system to estimate pose even when feature points are not visible in the current camera view by using spatial relationships from the 3D model.
Data Source
AI summary
A navigation program for an autonomous vehicle, the navigation program configured to: receive an initial model of an object to be inspected by the autonomous vehicle; identify an inspection target associated with the initial model of the object; and determine an inspection location for the autonomous vehicle from which inspection target is inspectable by an inspection system of the autonomous vehicle, wherein the initial model includes one or more convex shapes representing the object.