Behavior Predictability Scoring for Autonomous Route Changes
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Autonomous vehicles face challenges in predicting behaviors that are predictable to other road users, which can lead to unsafe navigation decisions when changing routes or navigating through environments with uncertain expectations.
Innovation Solution
A behavior prediction model is used to generate a predictability score for candidate future behaviors, comparing them with predicted future behaviors from the perspective of other road users, determining whether the behavior is consistent and safe based on route, action type, and turn consistency scores.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Adaptability or versatility
If autonomous vehicles use full-vehicle predictions for navigation decisions, then the vehicle can plan routes more effectively, but the predictability of behavior to other road users decreases
Solution Approach 1:
The system segments the navigation decision-making process into two distinct components: (1) a behavior prediction model that generates predictions based on observable scene information, and (2) a predictability evaluation module that assesses how well candidate behaviors align with these predictions. This segmentation allows the vehicle to maintain effective navigation planning while ensuring predictability by evaluating candidate routes against the prediction model before execution.
Solution Approach 2:
The system implements a feedback mechanism where the behavior prediction model continuously generates predictions about vehicle behavior from the perspective of other road users. These predictions serve as feedback to evaluate whether candidate navigation decisions are predictable, allowing the system to adjust its planning to maintain both effectiveness and predictability.
2Adaptability or versatility
If the vehicle changes navigation plan in response to reroute requests, then routing flexibility improves, but safety decreases due to unpredictable behavior
Solution Approach 1:
The system performs preliminary evaluation of candidate future behaviors using the behavior prediction model before actually executing navigation changes. By assessing predictability in advance, the system can approve routing changes that maintain safety or reject those that would create unpredictable and potentially harmful situations, thus enabling flexible routing while mitigating safety risks.
Solution Approach 2:
The predictability evaluation acts as a preliminary anti-action mechanism that prevents unsafe navigation changes before they occur. By evaluating whether candidate behaviors align with predictions from other road users' perspectives, the system blocks potentially harmful unpredictable actions while allowing safe flexible routing decisions.
Data Source
AI summary
Methods, systems, and apparatus, including computer programs encoded on a computer storage medium, for predictability estimation using a behavior prediction model. One of the methods includes receiving a candidate future behavior to be performed by an agent in an environment after a current time point; receiving data characterizing a scene that includes the agent in the environment as of the current time point; processing a behavior prediction input generated from the data using a behavior prediction model, wherein the behavior prediction model is configured to receive the behavior prediction input and to process the behavior prediction input to generate a behavior prediction output that characterizes a set of predicted future behaviors for the agent after the current time point; and determining a predictability score for the candidate future behavior by comparing the candidate future behavior with the behavior prediction output.