International Journal of Scientific Research and Engineering Development

Safety and cost-effective testing of autonomous vehicles (AVs) is a challenging task in transportation engineering. On-road testing is unable to test all risky scenarios that may occur with a self-driving vehicle, and manually building simulation libraries is time-consuming and expensive. This paper describes the Synthetic Intelligence-Driven Simulation Framework (SIDSF), which automatically generates different and realistic scenarios in a mixed-reality environment using generative AI. SIDSF uses naturalistic traffic, sensor noise, bad weather, and random pedestrian behaviour - all with statistical properties that resemble real-world conditions. SIDSF consists of four components: a scenario generator, a behavioural model for traffic, a physics-based sensor simulator and an ego vehicle controller that learns through reinforcement learning. Preliminary experiments demonstrate the benefits of higher rare-event coverage, reduced policy training time and fewer collisions compared to traditional simulators. This suggests that the use of simulation created by AI, using maps that are digital twins and feedback that is a mixed reality, is a viable and more efficient way to validate AV.

Sameer Kumar Sahu, Shivam Goyal,"Synthetic Intelligence–Driven Simulation Framework for Autonomous Mobility in Mixed-Reality Vehicular Environments" International Journal of Scientific Research and Engineering Development, V9(3): Page(307-311) May-June 2026. ISSN: 2581-7175. www.ijsred.com. Published by Scientific and Academic Research Publishing.