The CARLA descendant that flies drones and walks pedestrians

What it does HUTB is a cinematic-grade physics simulator for training and validating algorithms across autonomous vehicles, UAVs, and embodied humans. It ships with open digital assets—towns, buildings, vehicles, pedestrians—and exposes sensor simulation, traffic management, and multi-physics engines through a Python API. The target is robotics and self-driving research, not entertainment.

The interesting bit The project is refreshingly candid about its genealogy: it is explicitly built atop CARLA, AirSim, DReyeVR, and OpenSim, welding those ecosystems into a single package that adds VR cockpit driving, drone flight modes, and biomechanical pedestrian movement. It also includes China-specific map assets such as Hunan University of Technology and Business and the Changsha software park.

Key highlights

• Triple-mode support: autonomous driving, UAV simulation, and VR human-in-the-loop control via steering wheel or keyboard
• Bundles free digital assets and offers a no-compile Unreal Editor drop for artists
• Hooks into ROS, ScenarioRunner, SUMO/PTV Vissim, and Nvidia tools (SimReady, Cosmos)
• Demands serious hardware: NVIDIA RTX 2070 or better and 16 GB RAM minimum
• Python API covers versions 3.7 through 3.14

Caveats

• The codebase is integration-heavy, standing on the shoulders of CARLA, AirSim, and OpenSim rather than building a new engine from scratch
• Steep hardware floor and Unreal Engine overhead relegate it to workstation-class machines
• Documentation and primary interface are Chinese-first, which may add friction for international users

Verdict Worth evaluating if you need a unified CARLA-derived sandbox that also flies drones and walks biomechanically informed pedestrians. Skip it if you want a lightweight, original simulator or lack the GPU muscle.