Autonomous Vehicle Simulation Calibration Using Road Sensor Data
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Conventional methods for calibrating autonomous vehicle models heavily rely on extensive road tests, which are labor-intensive and risky, especially for vehicles like semi-trailer trucks experiencing unusual conditions, and lack efficient simulation techniques to accurately simulate driving behaviors and road dynamics.
Innovation Solution
A simulation framework using sensor data from prior road tests to calibrate longitudinal and lateral dynamic-related parameters of autonomous vehicle models, employing filtering and estimation techniques to determine road slope and banking angle, and adjust vehicle parameters based on yaw rate and throttle differences, allowing for model calibration without real-world testing.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If extensive road tests are performed to calibrate autonomous vehicle models, then measurement precision of vehicle dynamics parameters is improved, but loss of time and productivity deteriorate due to labor-intensive testing
Solution Approach 1:
The system performs preliminary calibration by simulating driving behaviors using sensor data from prior road tests before actual calibration is needed. The simulation framework pre-computes expected vehicle responses under various conditions, storing these results for later comparison and calibration adjustment, thereby reducing the need for extensive repeated road testing.
Solution Approach 2:
The patent creates a virtual copy of the autonomous vehicle's driving behavior through simulation. Instead of repeatedly testing the physical vehicle on roads, the system replicates driving scenarios in a simulated environment using sensor data from previous tests, allowing calibration to be performed on the virtual model rather than requiring extensive physical road tests.
2Manufacturing precision
If extensive road tests are conducted to calibrate vehicle models, then manufacturing precision of autonomous driving systems is improved, but device complexity increases due to extensive testing requirements
Solution Approach 1:
The simulation framework acts as an intermediary between raw sensor data and the autonomous vehicle model calibration process. Instead of directly complex road testing, the system uses sensor data to drive simulations that mediate the calibration process, simplifying the overall system architecture by replacing complex physical testing infrastructure with computational simulation.
Solution Approach 2:
The patent replaces the mechanical road testing system with a computational simulation system. Instead of physically testing the vehicle on roads with all associated complexity, the system substitutes mechanical testing with software-based simulation that uses sensor data to replicate driving behaviors and calibrate models computationally.
3Measurement precision
If road tests are performed to obtain sensor data for calibration, then measurement precision of driving behaviors is improved, but object-affected harmful factors increase due to safety risks during testing
Solution Approach 1:
The system performs preliminary data collection during normal operation or prior controlled tests, storing sensor data for later simulation and calibration. By collecting data beforehand and using it in simulated environments for calibration, the system eliminates the need to perform additional risky road tests solely for calibration purposes.
Solution Approach 2:
The patent uses sensor data from previous tests to create simulated driving scenarios that replicate real-world conditions without requiring actual road tests. The simulation copies the essential characteristics of real driving behaviors, allowing calibration to be performed in a safe virtual environment while maintaining measurement precision.
Data Source
AI summary
Vehicle dynamics related parameter(s) of an autonomous vehicle model can be calibrated so that the autonomous vehicle model can more accurately determine the driving related behaviors of the autonomous vehicle. An example method comprises obtaining, from sensor data, vehicle related parameters; performing a first determination of a slope of a road and a banking angle of the road based on a pitch angle of the vehicle and a roll angle of the vehicle, respectively; performing a second determination of a set of parameters that describe a driving-related operation of the vehicle; performing a third determination that at least one difference between at least one value from the set of parameters and a corresponding parameter from the plurality of vehicle related parameters exceeds at least one threshold value; and obtaining, in response to the third determination, a calibrated longitudinal dynamic-related parameter or a calibrated lateral dynamic-related parameter.