Autonomous Driving Path Planning With Generative Guide Information
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Data-driven autonomous driving methods struggle to recognize undefined objects and identify the intentions of pedestrians and surrounding vehicles, particularly in novel or unlearned driving situations, limiting their ability to respond effectively to changing environments.
Innovation Solution
A knowledge-driven autonomous driving method that utilizes a generative neural network model to generate guide information based on encoded sensor data, incorporating past driving experience and accumulated knowledge to enhance recognition and decision-making in various driving scenarios.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Productivity
If data-driven artificial neural network models are used for autonomous driving, then the system can process large amounts of driving data and make decisions based on learned patterns, but the system fails to recognize undefined objects and identify intentions in novel driving situations
Solution Approach 1:
The patent introduces a world model as an intermediary component that generates hypothetical driving scenarios and their outcomes. This world model acts as a mediator between the data-driven neural network and the decision-making process, allowing the system to simulate and reason about novel situations before making actual driving decisions. The world model enables the system to handle undefined objects and intentions by creating virtual representations of potential scenarios.
Solution Approach 2:
The system performs preliminary actions by generating multiple hypothetical driving scenarios and evaluating their outcomes before executing actual driving maneuvers. The world model pre-simulates various possible situations and their consequences, allowing the autonomous driving system to prepare decision-making frameworks in advance for novel situations that may arise during actual operation.
2Productivity
If the autonomous driving system uses traditional path determination methods, then the system can operate efficiently with established driving patterns, but the system cannot effectively respond to new and changing driving environments
Solution Approach 1:
The patent implements a dynamic system where the world model continuously generates and updates hypothetical scenarios based on current driving conditions. The path determination process is made dynamic by allowing the system to adaptively select and evaluate different driving paths through simulated scenarios, enabling efficient response to changing environments while maintaining operational efficiency through structured evaluation frameworks.
Solution Approach 2:
The system changes parameters by modifying the hypothetical scenarios and outcomes generated by the world model based on detected driving conditions. When novel situations are detected, the system adjusts the parameters of simulated scenarios to reflect new environmental conditions, allowing flexible adaptation while maintaining efficient operation through parameter-based control of the decision-making process.
3Speed
If the system generates guide information at high frequency, then the path determination responds quickly to changing conditions, but the computational load and processing time increase
Solution Approach 1:
The patent applies partial action by generating guide information at selectively adjusted frequencies based on the novelty and complexity of detected driving situations. For routine situations, the system reduces the frequency of world model updates and guide information generation, while increasing frequency only when novel conditions are detected. This allows quick response when necessary while reducing unnecessary computational overhead during normal operation.
Data Source
AI summary
A method of determining a final path of a moving object includes acquiring pieces of sensor data, determining a first path of the moving object based on the pieces of sensor data, inputting at least one piece of sensor data among the pieces of sensor data into an encoder that encodes the at least one piece of sensor data, inputting the encoded at least one piece of sensor data into a generative neural network model that generates guide information on a path of the moving object, and determining the final path of the moving object, based on the first path and the guide information.