Autonomous Vehicle Lane Relationship Classification for Simulation Training
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Current motion planning systems for autonomous vehicles face challenges in efficiently training models on relevant simulation scenarios due to the limitless and often irrelevant possibilities, leading to inefficient use of resources and time, necessitating methods to identify and develop effective simulation scenarios.
Innovation Solution
A method and system for classifying interactions in simulation scenarios, including identifying intersections and classifying vehicle-actor interactions based on lane relationships, path directions, and semantic information, to prioritize and rank scenarios for training, focusing on diverse and relevant interactions such as moving in the same or opposing directions, and actors crossing from the left or right.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Quantity of substance
If purely random simulation scenarios are used for training, then the motion planning model can be trained on a large number of scenarios, but the computing resources and time are wasted on irrelevant or extremely unlikely events
Solution Approach 1:
The patent changes the parameter of scenario selection from random to classified based on lane relationships. By introducing classification parameters (same direction, opposing directions, crossing from left, crossing from right), the system filters simulation scenarios to focus on relevant interactions, reducing wasted computing resources while maintaining adequate training data quantity
Solution Approach 2:
The patent extracts and isolates relevant simulation scenarios from the vast space of possible scenarios by applying classification filters. Only scenarios containing specific lane relationship interactions are selected for training, removing irrelevant scenarios from the training set to optimize resource utilization
2Reliability
If manual development of unique simulation scenarios is performed, then the motion planning model can be trained on relevant scenarios, but significant investment in time and manpower is required
Solution Approach 1:
The system performs self-service by automatically classifying simulation scenarios based on lane relationships without requiring manual curation. The classification is performed programmatically using the planned path of the vehicle and lane occupancy information of actors, eliminating the need for manual scenario development while maintaining high relevance
Solution Approach 2:
The classification system serves multiple functions simultaneously: it organizes scenarios by lane relationship type, identifies relevant interactions for training, and provides a framework for filtering scenarios. This multi-functional approach replaces multiple manual processes with a single automated system
Data Source
AI summary
Disclosed herein are system, method, and computer program product embodiments for assigning a lane relationship between an autonomous vehicle and other actors near an intersection. For example, the method includes executing a simulation scenario that includes features of a scene through which a vehicle may travel, the simulation scenario including one or more actors. The method further includes identifying an intersection between a first road and a second road in the simulation scenario, wherein the intersection is in a planned path of the vehicle. In response to one of the actors occupying a lane of either the first road or the second road, the method includes classifying the interaction between the vehicle and the actor based on the intersection, the path of the vehicle, and the lane occupied by the actor.