Arena simulation platform designed to accelerate Gatik autonomous trucking - The Robot Report

Gatik AI Inc. has launched Arena, a new simulation platform designed to accelerate the development and validation of its autonomous vehicles (AVs) by generating structured, controllable synthetic data. Arena addresses the limitations of traditional real-world data collection, which is time-consuming, expensive, and often unsafe, especially when capturing rare or high-risk scenarios. The platform uses advanced AI techniques such as neural radiance fields (NeRFs), 3D Gaussian splatting, and diffusion models to create photorealistic, high-fidelity simulations from various data inputs like segmentation maps, lidar, and HD maps. This enables comprehensive, closed-loop testing of the full autonomy stack, including multiple sensors (cameras, lidar, radar) and vehicle dynamics, while allowing scenario editing and A/B testing to simulate diverse environmental and traffic conditions. Arena aims to reduce the sim-to-real gap significantly, providing synthetic data that is sufficient for Gatik’s safety case and machine learning workflows without heavy reliance on annotated real-world data.