Occluded Object Emergence Prediction for Vehicle Path Planning
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Planning systems in autonomous and semi-autonomous vehicles face computational challenges in predicting numerous possible behaviors for dynamic objects in environments with occluded regions, making it difficult to determine potential interactions and plan safe maneuvers.
Innovation Solution
A vehicle computing device employs a model to analyze sensor data and generate a discretized representation of the environment, predicting the likelihood of objects in occluded regions and their potential impact, allowing for reduced computational load and improved safety by planning for unexpected object emergence.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Reliability
If planning systems predict numerous possible behaviors for each detected dynamic object, then vehicle safety is improved, but computational cost becomes excessive and may be impossible with onboard computing capabilities
Solution Approach 1:
The patent segments the environment into occluded and non-occluded regions, and further segments dynamic objects into those in occluded regions versus those in non-occluded regions. This segmentation allows the system to apply different prediction strategies: full behavior prediction for non-occluded objects and simplified emergence probability calculation for occluded objects, thereby reducing overall computational load while maintaining safety for critical objects.
Solution Approach 2:
The patent applies partial action by performing comprehensive behavior prediction only for non-occluded objects, while using a simplified approach (calculating emergence probability based on object type and occlusion characteristics) for occluded objects. This selective application of prediction depth reduces computational expenditure while still addressing safety concerns for the most critical visible objects.
2Measurement precision
If the system performs comprehensive predictions for all dynamic objects including those in occluded regions, then prediction accuracy is improved, but the complexity of the planning system increases
Solution Approach 1:
The planning system is segmented into different processing paths: one for non-occluded objects using full behavior prediction models, and another for occluded objects using emergence probability calculations. This segmentation reduces overall system complexity by applying simpler logic to occluded objects where full prediction would be computationally prohibitive and less reliable due to sensor occlusion.
Solution Approach 2:
The patent applies local quality by tailoring the prediction approach to the specific characteristics of each object's location. For occluded objects, the system calculates emergence probability based on object type and occlusion properties rather than performing full behavior prediction. This localized adaptation of prediction complexity optimizes the balance between accuracy and system complexity.
3Productivity
If the system reduces computational resources used for predicting occluded objects, then processing efficiency is improved, but the ability to detect and respond to emerging objects may be reduced
Solution Approach 1:
The system uses the characteristics of the occluded objects themselves (object type, size, and occlusion properties) to calculate emergence probability, rather than relying on complex sensor data processing. This self-service approach allows the system to maintain detection capability for occluded objects using minimal computational resources, as the prediction is based on inherent object properties rather than real-time sensor analysis.
Solution Approach 2:
The patent performs preliminary classification of objects into occluded and non-occluded categories, and pre-calculates emergence probabilities for occluded objects based on their type and characteristics. This preliminary action allows the system to prepare prediction data in advance with minimal computational effort, maintaining detection reliability while improving processing efficiency during real-time operation.
Data Source
AI summary
A vehicle computing device may implement techniques to predict behavior of objects or predicted objects in an environment. The techniques may include using a model to determine whether a potential object will emerge from an occluded region in the environment. The model may be configured to use one or more algorithms, classifiers, and/or computational resources to predict an intersection point and/or an intersection time between the potential object and the vehicle. Based on the predicted intersection point and/or the predicted intersection time, the vehicle computing device may control operation of the vehicle.