Open Vehicle Door Detection for Real-Time Autonomous Navigation
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Autonomous vehicles face challenges in navigating efficiently due to the vast amounts of data required for processing and storage, particularly in interpreting visual information and updating maps, which can lead to limitations in navigation accuracy and safety.
Innovation Solution
A system that includes a processor with circuitry and memory, capable of receiving image frames, identifying orientation indicators, and determining navigational actions based on the presence of open vehicle doors, using an open door detection network to guide the vehicle's actuators for navigation.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Reliability
If traditional mapping technology is used to navigate, then navigation coverage is improved, but data storage and processing burden increases significantly
Solution Approach 1:
The patent segments the navigation problem into two parts: using sparse map data for general navigation guidance and using computer vision techniques for real-time obstacle detection. This segmentation allows the system to rely on minimal map data while handling dynamic obstacles through image processing, thereby reducing overall data requirements while maintaining navigation accuracy.
Solution Approach 2:
The patent introduces computer vision algorithms as an intermediary between the sparse map data and the navigation decision-making process. This intermediary layer processes real-time visual information to detect obstacles and validate navigation paths, reducing dependence on comprehensive map data while ensuring safe navigation.
2Reliability
If comprehensive visual information is processed, then navigation safety is improved, but processing time and computational resources increase
Solution Approach 1:
The patent applies preliminary action by pre-training deep learning models offline to recognize obstacles and navigate complex scenarios. During real-time operation, these pre-trained models quickly process visual input without requiring extensive computational resources, thus maintaining high navigation safety while minimizing processing time.
Solution Approach 2:
The patent implements partial action by focusing computational resources on detecting and responding to critical obstacles rather than processing all visual information equally. The system identifies and prioritizes relevant visual features that impact navigation safety, reducing overall processing time while maintaining adequate safety monitoring.
Data Source
AI summary
A computer-implemented method for navigating a host vehicle may include receiving an image frame acquired by an image capture device associated with the host vehicle; identifying in the image frame a representation of a target vehicle; determining an orientation indicator associated with the target vehicle; based on the determined orientation indicator, identifying a candidate region of the acquired image frame where a representation of a vehicle door of the target vehicle is expected in an open door condition; extracting the candidate region from the image frame; providing the candidate region to an open door detection network; determining a navigational action for the host vehicle in response to an indication from the open door detection network that the candidate region includes a representation of a door of the target vehicle in an open condition; and causing an actuator associated with the host vehicle to implement the navigational action.