Controller Test Data with Plausible Initial Sensor Conditions
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing methods for validating Advanced Driver Assistance Systems (ADAS) and Autonomous Driving (AD) components using recorded real-world data often fail due to implausibilities in the initial state of simulated scenarios, leading to test failures.
Innovation Solution
A method is developed to process raw data from real-world sensors into test data by creating an 'introduction scenario' with virtual objects, ensuring a plausible initial state and seamless transition to the original scenario, thereby avoiding implausible states that could cause test failures.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Reliability
If recorded real-world data is used directly for testing, then realism and fidelity are improved, but implausibilities in initial state cause test failures
Solution Approach 1:
The patent applies preliminary action by generating an introduction scenario before the actual test scenario. This introduction scenario creates a plausible initial state with virtual objects that gradually transition into the real recorded scenario, ensuring the control unit is properly primed and avoiding test failures due to implausible initial conditions.
Solution Approach 2:
The patent uses an introduction scenario as an intermediary between the test environment and the real recorded scenario. This intermediary contains virtual objects that serve as a bridge, allowing the control unit to gradually adapt from a controlled virtual environment to the complexity of real-world data, thereby resolving the contradiction between realism and initial state plausibility.
2Productivity
If simulation starts with sensor traveling at full speed, then test efficiency is improved, but implausible initial state causes test abort
Solution Approach 1:
The patent applies preliminary action by creating an introduction scenario that starts with the sensor at zero speed and gradually accelerates to the target speed. This preliminary acceleration phase ensures plausible motion transitions and prevents test aborts, while still allowing the main test to run efficiently at full speed.
Solution Approach 2:
The patent applies dynamics by making the sensor speed variable during the introduction scenario. Instead of starting at full speed, the sensor speed dynamically changes from zero to the target speed, creating realistic motion profiles that prevent test failures while maintaining test efficiency in the subsequent full-speed phase.
3Device complexity
If objects suddenly appear or disappear in simulation, then scenario complexity is reduced, but realism and continuity are worsened
Solution Approach 1:
The patent applies continuity of useful action by ensuring that objects in the introduction scenario appear and disappear gradually rather than suddenly. Virtual objects are introduced and removed in a continuous manner that matches realistic temporal patterns, maintaining both scenario manageability and temporal continuity.
Solution Approach 2:
The patent applies preliminary action by pre-planning the temporal introduction and removal of virtual objects in the introduction scenario. Objects are introduced gradually with appropriate timing and spacing, ensuring continuous and plausible scene evolution without sudden appearances or disappearances that would break realism.
Data Source
Figure 1
Figure 2
Figure 3
AI summary
The invention relates to a method for processing raw real-world data into test data for stimulating a control unit, comprising the following process steps: providing recorded raw real-world data, identifying real objects detected by a sensor from the raw data and creating path data sets, each describing a scene with images of these real objects at successive time points, providing a library of synthetic objects, assigning synthetic objects to detected images of the real objects, replacing the detected images with synthetic objects, and supplementing the temporally successive path data sets with supplementary data sets before the first path data set, such that the sensor has an absolute velocity of zero at time zero of the induction period and the temporally successive supplementary data sets contain synthetic objects.which show a quasi-continuous temporal sequence of movements that quasi-continuously lead to the synthetic objects of the first path data set at time zero of the recording duration, and generate the test data by converting the path data sets supplemented by the additional data sets into such raw data as would have been recorded by the sensor in the real world during the initiation and recording durations if the sensor had detected the synthetic objects as real objects. In this way, a method is provided for generating such test data that does not exhibit any implausibilities.