Autonomous Vehicle Maneuver Ordering With Awareness Graphs
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Conventional interaction systems in autonomous vehicles are inefficient as they consider every potential intersection, are time-dependent, and fail to account for the effects of the autonomous vehicle's actions on surrounding vehicles, leading to inefficient use of computing resources and inability to handle aberrant behaviors.
Innovation Solution
The autonomous vehicle generates an awareness graph based on sensor signals to determine the ordering of maneuvers for itself and surrounding vehicles, using nodes and directed edges to represent dependencies between maneuvers, allowing it to selectively focus on relevant intersections and account for its own actions on other vehicles.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Reliability
If conventional interaction systems analyze every potential intersection between the autonomous vehicle and surrounding vehicles, then completeness of intersection analysis is improved, but computational efficiency deteriorates
Solution Approach 1:
The patent extracts only the relevant intersections from the set of all potential intersections by using the awareness graph to identify which intersections involve vehicles that are actually aware of and responding to the autonomous vehicle's maneuvers. This eliminates computation on irrelevant intersections while maintaining analysis of all meaningful interactions.
Solution Approach 2:
The patent segments the intersection analysis into two distinct phases: (1) identifying relevant intersections using the awareness graph based on vehicle awareness and maneuver dependencies, and (2) analyzing only those segmented relevant intersections. This segmentation allows the system to maintain completeness for important cases while improving efficiency by excluding irrelevant cases.
2Device complexity
If conventional interaction systems assume the autonomous vehicle follows a particular path, then simplicity of path prediction is improved, but accuracy of predicting surrounding vehicles' responses deteriorates
Solution Approach 1:
The patent applies dynamics by making the path prediction adaptive rather than static. The awareness graph dynamically updates based on the autonomous vehicle's selected maneuvers, and surrounding vehicles' paths are re-predicted in response to these maneuvers. This allows the system to start with simple baseline predictions but refine them dynamically as maneuvers are selected and executed.
Solution Approach 2:
The patent implements feedback by using the awareness graph to capture how surrounding vehicles respond to the autonomous vehicle's maneuvers. The system predicts maneuvers, selects optimal ones, then updates the awareness graph based on actual responses, creating a closed-loop system that continuously improves prediction accuracy while maintaining computational tractability.
3Reliability
If conventional interaction systems resolve every intersection at each timestamp, then thoroughness of interaction resolution is improved, but use of computing resources deteriorates
Solution Approach 1:
The patent applies partial action by resolving only the subset of intersections that are relevant to the autonomous vehicle's current maneuvers and the surrounding vehicles' awareness states. Rather than exhaustively resolving every possible intersection, the system performs partial resolution focused on critical interactions identified through the awareness graph, thereby conserving computing resources while maintaining thoroughness for important cases.
Data Source
AI summary
An autonomous vehicle is described herein. The autonomous vehicle comprises sensor systems that generate sensor signals and a computing system. The computing system generates an awareness graph based upon the sensor signals. The awareness graph comprises nodes connected by directed edges. The nodes include a first node assigned to a first maneuver that is to be executed by a vehicle in a vicinity of the autonomous vehicle and a second node assigned to a second maneuver that is to be executed by the autonomous vehicle. The directed edges include a directed edge that extends from the second node to the first node indicating that execution of the second maneuver is dependent upon execution of the first maneuver. The autonomous vehicle determines an ordering of the first maneuver and the second maneuver based on the awareness graph and executes the second maneuver in accordance with the ordering.