AV Perception Object Relevance for Motion Planning Load Control
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Autonomous vehicles face computational intensity and resource constraints when detecting and predicting multiple actors in their environment, leading to potential omission of relevant actors or violation of real-time processing constraints.
Innovation Solution
The system determines relevance of detected objects by identifying a region of interest based on a look-ahead lane and classifying objects as priority relevant, using a processor to assess which objects may interact within a time horizon, thereby prioritizing computations and excluding non-relevant objects from motion planning operations.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Reliability
If the AV continuously collects and processes perception data for multiple detected actors, then the accuracy of motion planning is improved, but the computational load and processing time increase significantly
Solution Approach 1:
The patent segments the set of all detected actors into two distinct subsets: priority relevant objects and non-priority objects. This segmentation is based on spatial criteria (objects in the region of interest) and temporal criteria (objects that may interact within the time horizon). By dividing the computational task into two segments, the system processes only the critical subset with high computational resources while using simplified processing for the remaining objects, thus resolving the contradiction between comprehensive analysis and computational feasibility.
Solution Approach 2:
The patent extracts and isolates the most critical subset of objects (priority relevant objects) from the complete set of detected actors. This extraction is performed by applying filtering criteria that identify objects within the region of interest and those with potential interaction within the time horizon. The extracted subset is then processed through the full motion planning pipeline, while the remaining objects are handled with reduced processing, effectively removing the computational burden from non-critical objects while preserving planning accuracy for relevant ones.
2Productivity
If the AV makes binary decisions to prioritize only high-priority agents, then the processing time is reduced, but relevant actors may be omitted or processing constraints are violated
Solution Approach 1:
The patent applies preliminary action by pre-defining the region of interest based on the vehicle's current lane and look-ahead distance, and pre-identifying the time horizon for potential interactions. These preliminary spatial and temporal boundaries are established before the actual object detection and classification occurs. By having these criteria pre-established, the system can quickly classify objects during real-time operation without complex decision-making, thus maintaining high processing speed while ensuring that all potentially relevant actors within the pre-defined boundaries are captured.
Solution Approach 2:
The patent introduces dynamics by making the region of interest and time horizon adaptive rather than fixed. The region of interest dynamically adjusts based on the vehicle's current lane, speed, and predicted trajectory, while the time horizon adapts based on the vehicle's operating conditions and the detected objects' characteristics. This dynamic approach ensures that the system maintains high processing efficiency by focusing on relevant objects while adapting to changing environmental conditions, preventing both omission of relevant actors and violation of processing constraints.
Data Source
AI summary
Methods of determining relevance of objects that a vehicle detected are disclosed. A system will receive a data log of a run of the vehicle. The data log includes perception data captured by vehicle sensors during the run. The system will identify an interaction time, along with a look-ahead lane based on a lane in which the vehicle traveled during the run. The system will define a region of interest (ROI) that includes a lane segment within the look-ahead lane. The system will identify, from the perception data, objects that the vehicle detected within the ROI during the run. For each object, the system will determine a detectability value by measuring an amount of the object that the vehicle detected. The system will create a subset with only objects having at least a threshold detectability value, and it will classify any such object as a priority relevant object.