Autonomous Vehicle Navigation Using Real-Time Scene Template Alignment
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Conventional autonomous driving systems rely heavily on pre-recorded high-definition 3D maps, which are limited by outdated or inaccurate data and require continuous connectivity, restricting their ability to navigate unfamiliar routes or recognize temporary traffic signals and obstacles.
Innovation Solution
A system and method for navigating autonomous driving vehicles by capturing and analyzing environmental data using a sensor assembly and computing device, which processes data to align a real-time 3D scene with a predefined template, allowing for autonomous navigation without relying on pre-recorded maps, utilizing sensors like lidar, cameras, and computer vision algorithms.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If conventional autonomous driving systems use pre-recorded high-definition 3D maps, then navigation accuracy is improved, but the system becomes limited by outdated data and requires continuous connectivity
Solution Approach 1:
The system performs preliminary actions by pre-defining scene templates that represent various road environments and traffic scenarios. These templates are prepared in advance and stored in the computing device, enabling the vehicle to quickly match and navigate unfamiliar environments without relying on pre-recorded maps or continuous cloud connectivity.
Solution Approach 2:
The system creates simplified copies of road scenes through predefined templates that capture essential geometric and semantic features. Instead of using complete high-definition maps, the system uses template copies that represent typical road configurations, allowing navigation in unfamiliar areas while maintaining computational efficiency and adaptability.
2Loss of information
If the system relies on pre-recorded HD 3D maps, then detailed information on signage and traffic lights is available, but the system cannot recognize temporary traffic signals or navigate through parking lots
Solution Approach 1:
The system transitions from static pre-recorded map data to dynamic real-time scene analysis. By continuously capturing and matching live camera images against predefined scene templates, the system can detect and respond to temporary traffic signals, construction zones, and changing road conditions that are not present in static maps.
Solution Approach 2:
The system implements feedback by continuously comparing real-time sensor data with predefined scene templates and adjusting navigation decisions accordingly. This real-time feedback loop enables the vehicle to recognize temporary traffic signals and adapt to changing environments, overcoming the limitations of static pre-recorded maps.
3Manufacturing precision
If high-definition maps are prepared using hyper-accurate mapping functionality, then centimeter scale precision is achieved, but the cost and time requirements increase significantly
Solution Approach 1:
The system replaces expensive, complex hyper-accurate mapping systems with more affordable sensor assemblies and predefined templates. Instead of investing in costly pre-mapping infrastructure, the system uses standard sensors combined with algorithmic template matching, achieving sufficient navigation precision without the high costs and complexity of professional mapping systems.
Solution Approach 2:
The system substitutes mechanical mapping systems (physical survey equipment, specialized vehicles) with computational methods. By replacing the mechanical pre-mapping process with software-based scene template matching and real-time image processing, the system achieves comparable navigation accuracy with significantly reduced hardware complexity and cost.
Applied Scientific Principles
This section explains which scientific principles are used to turn an abstract innovation direction into a practical engineering solution.
Function Achieved in This Case
Enables fully autonomous driving without the need for high-definition maps, allowing the vehicle to recognize and navigate through unfamiliar environments, detect obstacles, and perform safety-critical functions like lane recognition and traffic sign reading, mimicking human scene understanding.
Implementation Method 1
utilizing sensors like lidar, cameras, and computer vision algorithms
Implementation Method 2
capturing and analyzing information of global scene and local objects around the autonomous driving vehicle
Data Source
AI summary
A system and method for navigating an autonomous driving vehicle (ADV) by capturing and analyzing information of a global scene and local objects around the ADV, is disclosed. The system comprises a sensor assembly incorporated on the ADV and a computing device in communication with the sensor assembly. The sensor assembly is configured to collect environmental data around the ADV. The computing device comprises a processor, and a memory unit for storing a predefined scene template and environmental data. The computing device is configured to process the environmental data to identify a moving and static object. The computing device is further configured to observe an environmental scene around the ADV. The observed environmental scene is aligned with a predefined scene template. Further, the predefined scene template is adjusted using the processed environmental data. The computing device provides instruction to control the vehicle based on the adjusted scene template.