Vehicle Pose Estimation Using Depth Cameras in GPS-Denied Navigation
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Autonomous vehicles face navigation challenges in environments without satellite-based positioning signals, such as parking garages, where available localization data is inadequate, leading to cessation of autonomous operation.
Innovation Solution
An imaging system utilizing electronic depth cameras and a deep neural network to determine a vehicle's multi-degree of freedom pose, enabling navigation by capturing RGB and depth images, identifying keypoints, and constraining the pose to a ground plane, allowing the vehicle to navigate through structured environments without GPS data.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If GPS data is used for vehicle navigation, then navigation accuracy is improved, but the system fails in environments with weak or absent satellite signals
Solution Approach 1:
The patent introduces an imaging system with electronic depth cameras as an intermediary to capture visual information and determine vehicle pose when GPS is unavailable. The system uses cameras to detect keypoints on the vehicle and calculate multi-degree-of-freedom pose, serving as a mediator between the vehicle's navigation needs and the absence of satellite signals.
Solution Approach 2:
The navigation system is enhanced with multi-functionality by integrating both GPS-based navigation and vision-based pose estimation. The system can automatically switch between GPS mode and camera-based MDF pose determination, allowing it to function universally across different environmental conditions including areas with weak or no satellite coverage.
2Ease of manufacture
If traditional GPS-based navigation is used, then the system is simple to implement, but autonomous operation ceases without GPS data
Solution Approach 1:
The system performs preliminary action by pre-training a deep neural network with labeled depth images containing fiducial markers before actual navigation. This pre-training phase prepares the neural network to accurately determine vehicle pose from camera images, ensuring the system is ready to maintain autonomous operation when GPS becomes unavailable.
Solution Approach 2:
The system implements feedback through continuous monitoring of GPS signal availability and automatic switching between navigation methods. When GPS data becomes unavailable, the system activates the camera-based MDF pose determination and provides feedback to maintain autonomous navigation, ensuring continuous reliable operation.
3Reliability
If MDF pose determination with neural networks is implemented, then navigation reliability in GPS-denied environments is improved, but device complexity increases
Solution Approach 1:
The system segments the navigation function into distinct modules: GPS-based navigation for open environments and camera-based MDF pose determination for GPS-denied environments. This segmentation allows each module to be optimized independently and simplifies the overall system architecture by clearly defining when each module should be activated.
Solution Approach 2:
The system uses copying by creating a virtual representation of the vehicle's pose and position based on camera images and neural network processing. The deep neural network generates MDF pose estimates that copy the essential navigation information needed for autonomous operation, reducing the need for complex hardware while maintaining reliability.
Data Source
AI summary
A computer, including a processor and a memory, the memory including instructions to be executed by the processor to capture, from a camera, one or more images, wherein the one or more images include at least a portion of a vehicle, receive a plurality of keypoints corresponding to markers on the vehicle and instantiate a virtual vehicle corresponding to the vehicle. The instructions include further instructions to determine rotational and translation parameters of the vehicle by matching a plurality of virtual keypoints to the plurality of keypoints and determine a multi-degree of freedom (MDF) pose of the vehicle based on the rotational and translation parameters.