Vehicle Environment Parameter Modeling From Real and Virtual Sensors
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing autonomous vehicle systems face challenges in accurately and efficiently determining environmental parameters due to the virtually infinite number of scenarios they may encounter, leading to reduced accuracy and safety concerns, particularly when using separate algorithms for road curvature and quality estimation, and end-to-end learning techniques lack verifiability.
Innovation Solution
A machine learning model trained using real-world and virtual sensor data, including ground truth data, to concurrently determine road parameters and object distances, which is applied to a motion planner for controlling the vehicle, reducing complexity and improving safety by utilizing cross-sensor dependencies.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Ease of manufacture
If separate algorithms are used for determining road curvature and estimating road quality, then the system can process specific parameters independently, but the overall accuracy is reduced due to high dependency among road information components
Solution Approach 1:
The patent combines multiple separate algorithms for determining road curvature, road quality, and other road parameters into a single integrated machine learning model. This model processes sensor data concurrently to generate multiple road parameters simultaneously, eliminating the need for separate independent algorithms and accounting for the high dependency among road information components, thereby improving overall accuracy.
2Productivity
If end-to-end learning is implemented for autonomous vehicle control, then the system can generate decisions directly from sensor inputs, but the technique lacks verifiability and becomes risky for safety-critical applications
Solution Approach 1:
The patent segments the autonomous vehicle system into distinct functional modules: a machine learning model for perception that generates environmental parameters from sensor data, and a separate motion planning component that uses these parameters to generate control decisions. This segmentation allows the perception system to leverage the efficiency of end-to-end learning while maintaining verifiability in the motion planning stage, where safety-critical decisions can be validated and controlled independently.
3Reliability
If procedural processing is used to handle sensor inputs, then the system can execute pre-programed operations reliably, but it cannot account for the large number of situations an autonomous vehicle can encounter
Solution Approach 1:
The patent employs a machine learning model that learns to map sensor inputs to environmental parameters through training on diverse datasets representing various driving situations. This allows the system to adapt to a wide range of scenarios by changing its internal parameters and decision boundaries based on learned patterns, rather than relying on fixed pre-programed operations, while maintaining reliability through supervised training with ground truth data.
Data Source
AI summary
First training sensor data detected by a plurality of real-world sensors are obtained. The first training sensor data is associated with physical environment conditions. Second training sensor data detected by a plurality of virtual sensors are obtained. The second training sensor data is associated with simulated physical conditions of a virtual environment. A machine learning model is trained using both real-world and virtual training datasets including the first training sensor data, the second training sensor data, and respective sensor setting parameters of the plurality of real-world sensors and the plurality of virtual sensors. The real-world and virtual training datasets used to train the machine learning model include indications associated with the respective sensor parameter settings including one or more of the following: different scan line settings or different exposure settings. The machine learning model is provided for use in generating current parameters of an environment in which a vehicle operates.