Vehicle Computing System Hyper Planning for Object Prediction
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Autonomous and semi-autonomous vehicle planning systems face computational challenges in processing large numbers of detected objects in their environment, leading to inefficient simulations and limited detail due to onboard computing limitations.
Innovation Solution
The implementation of different models for different objects and regions in the environment, allocating more computational resources to relevant objects based on features such as proximity, type, and likelihood of impact, allowing for more detailed prediction processing for critical objects while reducing resources for less relevant ones.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If computational operations are performed on each detected object, then object detection accuracy is improved, but computational load increases making it impossible with onboard computing capabilities
Solution Approach 1:
The patent segments the environment into multiple regions (first region, second region, third region) with different computational processing levels. Objects in different regions receive different amounts of computational resources based on their relevance to the vehicle, allowing accurate detection of critical objects while reducing processing for less important areas, thus resolving the contradiction between detection accuracy and computational load.
Solution Approach 2:
The patent applies different quality levels of computational processing to different spatial regions. The first region (highest priority) receives full computational processing for maximum detection accuracy, the second region receives moderate processing, and the third region receives minimal processing. This local differentiation allows the system to maintain high detection accuracy where needed while reducing overall computational load to fit onboard capabilities.
2Power
If simulations are performed with limited onboard computing capabilities, then computational load is reduced, but simulation detail and complexity are limited
Solution Approach 1:
The patent divides the simulation processing into region-based segments where the first region receives detailed simulations with high complexity, the second region receives moderate simulation detail, and the third region receives simplified simulations. This segmentation allows the system to perform comprehensive simulations on critical objects while using simplified models for less important objects, maintaining simulation detail where needed while managing overall computational load.
Solution Approach 2:
The patent applies different levels of simulation detail and complexity to different spatial regions. High-fidelity simulations are concentrated in the first region where objects have the greatest impact on vehicle operation, while lower-fidelity simulations are used in the second and third regions. This local quality approach ensures that computational resources are allocated to produce detailed simulations only where necessary, resolving the contradiction between computational load and simulation detail.
3Ease of operation
If equal computational resources are allocated to all objects, then object processing consistency is maintained, but computational efficiency decreases due to over-processing irrelevant objects
Solution Approach 1:
The patent segments objects into different categories based on their spatial region and relevance to the vehicle. Objects in the first region are processed with high priority and detailed analysis, objects in the second region receive moderate processing, and objects in the third region receive minimal processing. This segmentation replaces equal resource allocation with differentiated allocation based on actual need, maintaining consistency within each segment while dramatically improving overall computational efficiency by avoiding over-processing of irrelevant objects.
Solution Approach 2:
The patent applies different processing intensities to different spatial regions and object types. The first region (containing objects most relevant to vehicle operation) receives high-quality processing with detailed analysis, while the second and third regions receive progressively lower quality processing. This local quality approach maintains processing consistency for objects within the same relevance category while optimizing overall computational efficiency by adapting processing quality to the actual importance of each object.
Data Source
AI summary
A vehicle computing system may implement techniques to predict behavior of objects detected by a vehicle operating in the environment. The techniques may include determining a feature with respect to a detected objects (e.g., likelihood that the detected object will impact operation of the vehicle) and/or a location of the vehicle and determining based on the feature a model to use to predict behavior (e.g., estimated states) of proximate objects (e.g., the detected object). The model may be configured to use one or more algorithms, classifiers, and/or computational resources to predict the behavior. Different models may be used to predict behavior of different objects and/or regions in the environment. Each model may receive sensor data as an input, and output predicted behavior for the detected object. Based on the predicted behavior of the object, a vehicle computing system may control operation of the vehicle.