Inclusive Simulator for Diverse Synthetic Pedestrian Data Generation

One of the critical challenges in connected automated mobility (CCAM) and AI-based systems is the lack of representative datasets that reflect the broad range of vulnerable road users (VRUs) found in real-world scenarios, including underrepresented groups such as people with disabilities. This can lead to AI systems not equipped to handle the full diversity of VRUs, thus compromising the safety and inclusivity of CCAM.

The AWARE2ALL project, which aims to develop inclusive HMI (Human-Machine Interface) and safety systems for autonomous vehicles, is using DiverSim simulator to address dataset representation issues. This simulator (developed by Vicomtech) generates balanced datasets including in equal proportions a wide variety of genders, different ethnic minorities, individuals with various disabilities (from using wheelchairs to visual impairment). DiverSim ensures that no minority is underrepresented, thus avoiding biases that are common in most openly available datasets. AWARE2ALL will use the data produced to assist in re-training and validating AI models to better accommodate the needs of diverse individuals.

DiverSim simulator and its data generation tools are now open source; all assets and animations used are carefully selected with licenses allowing researchers and developers worldwide to use them freely for training and validating AI models designed to handle a wide range of pedestrian types and scenarios.

Source: More information available in the original article published here