Trajectory Scoring via Rulebook Metrics for Autonomous Vehicles
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Validating the behavior of autonomous vehicles in diverse driving scenarios is challenging due to the complexity of driving rules and their priority structures, particularly when rules need to be violated, such as in scenarios involving illegally parked cars.
Innovation Solution
A rule-based trajectory evaluation system that divides trajectories into time intervals, evaluates each interval for rule violations using a structured rulebook with violation metrics, and aggregates these metrics to generate a score for the entire trajectory, allowing for the identification of rule violations and their severity.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Adaptability or versatility
If a planner generates candidate trajectories considering multiple driving rules, then the vehicle can navigate complex scenarios, but it becomes difficult to validate which trajectory is correct when rules conflict
Solution Approach 1:
The patent segments the validation process into distinct components: a rule engine that evaluates individual rules, a priority structure that orders rules, and a trajectory scorer that aggregates results. This segmentation transforms the complex validation problem into manageable modular components that can be independently developed and tested.
Solution Approach 2:
The patent introduces an intermediary rule engine and priority structure that mediates between the planner's candidate trajectories and the validation process. This intermediary layer systematically evaluates each trajectory against ordered rules, providing a clear mechanism for resolving rule conflicts without requiring complex ad-hoc reasoning.
2Reliability
If driving rules are evaluated with strict compliance requirements, then safety is improved, but legitimate rule violations (e.g., overtaking parked cars) cannot be accommodated
Solution Approach 1:
The patent makes the rule evaluation dynamic by introducing a priority-ordered rule structure where rules can be selectively applied or overridden based on context. The system dynamically determines which rules to enforce and which to violate based on the aggregated score from multiple rule evaluations, allowing flexible adaptation to legitimate violation scenarios.
Solution Approach 2:
The patent changes the parameter of rule evaluation from binary (compliant/non-compliant) to a multi-dimensional scoring system. Each rule contributes a score based on its priority and violation severity, and the overall trajectory score aggregates these parameters. This parameter transformation allows the system to distinguish between acceptable and unacceptable rule violations.
3Measurement precision
If multiple candidate trajectories are generated for evaluation, then better path selection is achieved, but the validation process becomes more computationally intensive
Solution Approach 1:
The patent applies partial action by evaluating only the necessary rules for each trajectory rather than exhaustively checking all possible rules. The priority-ordered structure allows the system to stop evaluating once a trajectory is clearly superior or inferior, reducing computational effort while maintaining evaluation accuracy.
Solution Approach 2:
The patent performs preliminary action by pre-ordering rules in a priority structure and pre-defining violation metrics before trajectory evaluation begins. This preliminary setup enables efficient runtime evaluation of multiple trajectories without recalculating rule priorities or metrics, significantly reducing computational overhead during the validation process.
Data Source
AI summary
Enclosed are embodiments for scenario-based behavior specification and validation. In an embodiment, a method comprises: obtaining, using at least one processor, at least one trajectory associated with a driving scenario for an autonomous vehicle system; evaluating, using the at least one processor and at least one rulebook, the at least one trajectories to determine whether the at least one trajectory violates at least one rule in the at least one rulebook, wherein each rule in the rulebook is associated with at least one violation metric that is used to determine a degree to which the rule was satisfied or violated; determining, using the at least one processor and the at least one violation metric, a score for the at least one trajectory; and sending, using the at least one processor, the score to at least one of a software module in a software stack of the autonomous vehicle system, a simulation of the autonomous vehicle system or as a report or in a visual presented through a user interface of a cloud-based platform.