Autonomous Vehicle Motion Planning With Conditional Maneuver Restriction
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Autonomous vehicles face inefficiencies in motion planning due to the need to evaluate numerous potential maneuvers, including conditionally disallowed actions, which can lead to computationally expensive and unpredictable behaviors, especially in situations with multiple actors and dynamic road conditions.
Innovation Solution
A system and method for an autonomous vehicle to acquire data on actors in the roadway, predict conditionally disallowed actions, and automatically restrict them from the motion plan to prevent undesirable maneuvers by generating a compensating trajectory that aligns with predictable behaviors and reduces computational complexity.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Adaptability or versatility
If the autonomous vehicle evaluates numerous potential maneuvers including conditionally disallowed actions, then the vehicle can handle complex road situations, but the computational cost and processing time increase significantly
Solution Approach 1:
The system pre-identifies and restricts conditionally disallowed actions before they are evaluated during trajectory optimization. By predicting actor trajectories and determining which actions would be disallowed in advance, the system eliminates unnecessary computational evaluations, reducing processing time while maintaining the ability to handle complex situations through selective action restriction based on predicted road conditions
Solution Approach 2:
The motion planning process is segmented into distinct phases: actor trajectory prediction, conditionally disallowed action identification, and trajectory optimization. This segmentation allows the system to separately handle prediction and optimization tasks, evaluating only relevant maneuvers in the optimization phase rather than all possible maneuvers, thereby reducing computational complexity while maintaining adaptability
2Adaptability or versatility
If the autonomous vehicle considers all potential maneuvers unconditionally, then the vehicle has maximum maneuvering options, but the motion plan becomes unnecessarily large and requires more resources
Solution Approach 1:
The system performs preliminary restriction of conditionally disallowed actions based on predicted actor trajectories and road conditions before generating the full motion plan. This preliminary filtering reduces the size of the motion plan by eliminating maneuvers that would be disallowed under predicted conditions, while maintaining adaptability by allowing unrestricted maneuvers when conditions permit
Solution Approach 2:
The system applies different quality levels to different maneuvers by restricting only those actions that are conditionally disallowed based on local road conditions and actor predictions, while leaving other maneuvers unrestricted. This selective restriction reduces motion plan complexity locally without globally limiting maneuvering options, maintaining adaptability where needed while reducing unnecessary complexity
3Manufacturing precision
If the autonomous vehicle performs extensive calculations to optimize candidate trajectories, then the vehicle can select optimal paths, but the computational cost becomes expensive and sometimes computationally infeasible
Solution Approach 1:
The system extracts and removes conditionally disallowed actions from the set of candidate maneuvers before performing trajectory optimization. By taking out these restricted actions based on predicted road conditions and actor trajectories, the system reduces the number of trajectories that need to be evaluated and optimized, lowering computational energy cost while maintaining optimization precision for the remaining relevant maneuvers
Solution Approach 2:
The system performs preliminary identification and restriction of disallowed actions before the computationally expensive trajectory optimization phase. This preliminary action reduces the search space for optimization, allowing the system to achieve the same level of optimization precision with fewer computational resources by focusing only on permissible maneuvers
4Productivity
If the autonomous vehicle uses discretization or random sampling to limit maneuvers, then the computational load is reduced, but the system may eliminate necessary maneuvers that the vehicle must perform
Solution Approach 1:
The system changes the parameter of action availability by dynamically restricting or allowing maneuvers based on predicted road conditions and actor trajectories. Instead of using fixed discretization or random sampling, the system adjusts which maneuvers are available based on real-time predictions, ensuring necessary maneuvers are not eliminated while maintaining computational efficiency through condition-based restriction
Solution Approach 2:
The system implements dynamic maneuver restriction where the availability of maneuvers changes based on predicted road conditions and actor behaviors. This dynamic approach replaces static discretization methods, allowing the system to maintain high productivity by restricting only when necessary while ensuring reliability by allowing necessary maneuvers when predictions indicate they are safe and appropriate
Data Source
AI summary
Provided are systems, methods, and computer program products for controlling an autonomous vehicle (AV) to maneuver in a roadway, comprising acquiring, data associated with an actor detected on a route of the AV in the roadway for sensing a trajectory of the actor, predicting that the trajectory of the actor includes at least one characteristic that is associated with invoking a conditionally disallowed action in the AV, automatically restricting the conditionally disallowed action from a motion plan of the AV to prevent the AV from executing the conditionally disallowed action in response to detecting that one or more conditions are present in the roadway, issuing a command to control the AV on a candidate trajectory generated to prevent an option for the conditionally disallowed action.