Autonomous Vehicle Agent Impact Scoring for Trajectory Planning
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Autonomous vehicles face challenges in generating timely and accurate prediction data for agents in their vicinity due to limited computational resources, especially in busy environments where numerous agents are present, leading to insufficient prediction data for all agents.
Innovation Solution
An importance scoring model is implemented to determine relative impact scores of agents, allowing the system to concentrate computational resources on high-impact agents, using a more computationally efficient prediction model for low-impact agents, and generating training data through incremental processing techniques to improve agent impact estimation.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Reliability
If computational resources are allocated to generate prediction data for all agents in busy environments, then prediction coverage is improved, but computational resource exhaustion occurs
Solution Approach 1:
The patent segments agents into different priority levels based on their impact on vehicle trajectories. High-impact agents (those whose actions significantly affect planned trajectories) receive full computational resources for accurate prediction, while low-impact agents use simplified prediction models or are excluded from detailed analysis, thereby resolving the contradiction between comprehensive prediction coverage and computational resource limits
Solution Approach 2:
The patent applies different quality levels of prediction processing to different agents based on their local importance. Critical agents near the vehicle or with high maneuverability receive high-precision prediction resources, while distant or static agents receive minimal processing, optimizing the balance between prediction reliability and computational energy consumption
2Reliability
If comprehensive prediction data is generated for all agents, then navigation safety is improved, but processing time increases
Solution Approach 1:
The patent segments the agent population into priority categories and processes them in stages. High-priority agents are processed with full detail to ensure safety-critical decisions, while low-priority agents are processed more quickly or skipped, reducing overall processing time while maintaining navigation safety for critical scenarios
Solution Approach 2:
The patent implements partial processing by generating detailed prediction data only for a subset of high-impact agents rather than all agents. This partial action approach ensures sufficient prediction accuracy for safety-critical navigation decisions while dramatically reducing computation time by excluding low-impact agents from intensive processing
3Productivity
If importance scoring is implemented to prioritize agents, then computational efficiency is improved, but system complexity increases
Solution Approach 1:
The patent implements preliminary action by pre-computing importance scores for agents based on their spatial relationships and potential impact on vehicle trajectories. These pre-computed scores are stored and reused to quickly determine processing priorities without requiring complex real-time analysis, thereby improving computational efficiency while managing system complexity through advance preparation
Data Source
AI summary
Methods, systems, and apparatus, including computer programs encoded on a computer storage medium, for determining respective importance scores for a plurality of agents in a vicinity of an autonomous vehicle navigating through an environment. The respective importance scores characterize a relative impact of each agent on planned trajectories generated by a planning subsystem of the autonomous vehicle. In one aspect, a method comprises providing different states of an environment as input to the planning subsystem and obtaining as output from the planning subsystem corresponding planned trajectories. Importance scores for the one or more agents that are in one state but not in the other are determined based on a measure of difference between the planned trajectories.