Page 178 - AI for Good-Innovate for Impact Final Report 2024

AI for Good-Innovate for Impact

unseen scenarios. The virtual simulation platform can quickly generate large-scale data sets of
corner cases scenarios based on synthetic data.

At the same time, in order to reduce the accident rate in traffic scenarios, the intelligent driving
algorithm is connected to the virtual simulation platform, SIL software is implemented in the
loop, algorithm code is run and tested, and continuous iteration and optimization are carried
out, so that "traffic accidents stop in the virtual world".

40�2�2 Future work

The future work plan is divided into the following points:

1. Conduct big data analysis on the current open source data set, road test data and mass
production return data, and dig out the core corner cases in real scenarios;
2. Conduct 3D scene reconstruction and development for core corner cases scenes, edit
static and dynamic elements, and realize the generalization of scene contents; Further
improve the synthetic data set, including adding synthetic data for different scenarios,
weather conditions, road conditions, etc., to ensure data quality and diversity;
3. Based on the real data and synthetic data of intelligent driving, optimize the perception
algorithm of intelligent driving system, improve the recognition and understanding
ability of various complex scenarios, reduce the probability of misjudgment and missed
judgment, evaluate the performance and robustness of the perception algorithm, find
potential problems and make improvements;
4. Cooperate with industry, academia and government departments to jointly promote the
development and application of intelligent driving perception technology and promote
the application of synthetic data in the field of intelligent driving; The corner cases data
set is disclosed and the data conversion mode and method are set.

40�3� Use case requirements

• REQ-01: Simulation of different weather conditions: The synthetic data can simulate
different weather conditions, such as rainy, snowy, foggy, etc., to help the autonomous
driving system adapt to driving conditions in various weather conditions.
• REQ-02: Simulation of different road conditions: The synthetic data can simulate different
road conditions, such as highways, urban roads, rural roads, etc., to help the autonomous
driving system adapt to driving conditions in different road environments.
• REQ-03: Simulation of traffic situations: Synthetic data can simulate different traffic
situations, such as congestion, intersections, pedestrian crossing, etc., to help autonomous
driving systems adapt to complex traffic situations.
• REQ-04: Simulation of vehicle behavior: Synthetic data can simulate the behavior of other
vehicles, such as lane changes, overtaking, parking, etc., to help the autonomous driving
system predict the behavior of other vehicles and make corresponding driving decisions.

162