Hierarchical Edge Computing for Learned Vehicle Path Guidance
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Autonomous and semi-autonomous vehicles lack real-time access to prior navigation data, leading to suboptimal path planning and safety, as they approach environments without leveraging historical information from previous vehicles.
Innovation Solution
Implementing a hierarchical edge compute system that aggregates and compares actual vehicle paths and speeds with learned paths stored in edge servers, allowing for real-time adjustments and safety measures, such as modifying vehicle behavior or recomputing trajectories.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Loss of information
If autonomous vehicles use only local sensor data and onboard computing for path planning, then the system maintains simplicity and real-time responsiveness, but it cannot leverage historical navigation data from previous vehicles to improve safety and path optimization
Solution Approach 1:
The system divides the computing architecture into three hierarchical levels: onboard vehicle computers for real-time local processing, edge servers for regional data aggregation and learned path storage, and centralized data centers for comprehensive data management. This segmentation allows historical data to be leveraged without overwhelming any single computing component, resolving the contradiction between information access and system complexity.
Solution Approach 2:
Edge servers act as intermediaries between onboard vehicle systems and centralized data centers. They aggregate navigation data from multiple vehicles in a specific geographic region, store learned paths, and provide real-time comparisons to approaching vehicles. This intermediary layer enables historical data utilization while maintaining system modularity and managing complexity.
2Reliability
If autonomous vehicles aggregate and process navigation data from multiple vehicles in real-time, then path planning accuracy and safety improve, but computational load and data processing time increase
Solution Approach 1:
The system pre-computes and stores learned paths and probability distributions in edge servers before vehicles need them. As vehicles traverse a region, their navigation data is continuously aggregated and processed offline to build statistical models of optimal paths. When new vehicles enter the region, they immediately receive pre-computed guidance without requiring real-time aggregation, thus improving reliability without adding processing delays.
Solution Approach 2:
Edge servers are deployed in specific geographic regions to handle data aggregation and processing locally rather than requiring all vehicles to communicate with a centralized system. This distributes computational load across multiple edge locations, reducing the time required for data processing while maintaining high reliability through localized historical data utilization.
3Measurement precision
If the system stores and compares multiple learned paths for every road section, then navigation accuracy improves, but memory requirements and data storage complexity increase
Solution Approach 1:
Instead of storing complete detailed paths for every possible scenario, the system transforms navigation data into statistical parameters including probability distributions, expected trajectory models, and key decision point characteristics. This parameterization compresses large amounts of raw path data into concise mathematical representations that maintain high measurement precision while significantly reducing storage requirements.
Solution Approach 2:
The system stores learned paths and probability data selectively for road sections where historical data indicates significant variation or safety concerns. For routine, well-understood road sections, minimal baseline data is stored. This partial action approach ensures high precision is maintained where needed while avoiding unnecessary storage of redundant information, balancing accuracy with storage efficiency.
Data Source
AI summary
Embodiments of the present disclosure may include a method and system for navigating one or more environments using hierarchical edge computing. In some embodiments, the method may include sending location data to an edge server, where, in some embodiments, the location data may indicate a location of the ego-machine and path data that may indicate a planned path through a portion of an environment that may correspond to the ego-machine. Further, in some embodiments, the method may additionally include receiving a notification from the edge server associated with the planned path. In some embodiments, the notification is based on a comparison between the planned path to one or more learned paths. Additionally or alternatively, in some embodiments, the method may additionally include navigating the ego-machine through the portion of the environment based on the received notification.