Vehicle Trajectory Planning for Adverse Object Behavior
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Autonomous and semi-autonomous vehicle planning systems face computational challenges in predicting numerous possible behaviors of dynamic objects in complex environments, leading to inefficiencies and potential safety issues due to the high cost of computations and the possibility of overlooking adverse, unlikely object behaviors.
Innovation Solution
The implementation of an adverse prediction model that focuses on predicting adverse behaviors of objects, such as sudden changes in direction or speed, allowing the vehicle to plan for potential hazardous scenarios by determining candidate trajectories based on object characteristics and likelihoods, thereby improving safety and reducing computational load.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Reliability
If numerous predictions are performed for each detected dynamic object to ensure comprehensive safety coverage, then vehicle safety is improved, but computational cost increases and may exceed onboard computing capabilities
Solution Approach 1:
The patent extracts and focuses computational resources on predicting only adverse behaviors of dynamic objects rather than predicting all possible behaviors. The system identifies objects that may behave unexpectedly or change speed/direction quickly and devotes computational resources specifically to predicting their adverse actions, thereby reducing overall computational cost while maintaining safety for critical scenarios
Solution Approach 2:
The patent applies different prediction strategies to different objects based on their characteristics. High-maneuverability objects like bicycles and pedestrians receive adverse behavior predictions focused on sudden direction changes and speed variations, while other objects may use standard prediction methods. This localized approach optimizes computational resource allocation to where it is most needed for safety
2Measurement precision
If computational resources are devoted to predicting likely actions by objects, then prediction accuracy for common behaviors is improved, but adverse unlikely behaviors may be overlooked
Solution Approach 1:
The patent performs preliminary identification of objects that are capable of unexpected behaviors based on their physical characteristics (e.g., high maneuverability, ability to change speed quickly). Once identified, the system proactively predicts adverse behaviors for these objects in advance, ensuring that unlikely but dangerous actions are considered before they occur, rather than waiting for standard prediction methods to detect them
Solution Approach 2:
The patent changes the prediction parameters and models used for different object types. For objects with high maneuverability or unexpected behavior capability, the system uses adverse prediction models that specifically account for sudden direction changes, maximum acceleration, and erratic movements. This parameter change ensures that the prediction system is tuned to detect adverse behaviors rather than only predicting normal traffic flow patterns
Data Source
AI summary
A vehicle computer system implements techniques to predict behavior of objects detected by a vehicle operating in the environment. The techniques include using a model to determine a first object trajectory for an object (e.g., a predicted object trajectory) and/or a potential object in an occluded area, as well as a second object trajectory for the object or potential object (e.g., an adverse object trajectory). The model is configured to use one or more algorithms, classifiers, and/or computational resources to predict candidate trajectories for the vehicle based on at least one of the first object trajectory or the second object trajectory. Based on the predicted behavior of the object (or potential object) and the predicted candidate trajectories for the vehicle, a vehicle computer system controls operation of the vehicle.