Cyprian Kiplangat@mutaician

RoboHazard Arena is a research-based drone simulation project exploring how small vision-language models perform when placed inside a live 3D environment. The goal is not to give the model step-by-step instructions, but to give it a mission: reach the green target pad safely while avoiding obstacles. The drone can be controlled manually, or control can be delegated to Qwen3-VL. The model receives a camera view, telemetry, target information, obstacle information, and available actions, then chooses high-level movements such as move forward, strafe, ascend, descend, hover, or land. The project shows both the promise and limitations of small VLMs for embodied control. Qwen3-VL-2B succeeds in some simple cases but fails or hallucinates in others, proving that more research, better simulation, and stronger evaluation are needed before small models can reliably control small vehicles in dynamic real-world environments.

Grab your ideas, spin up compute, and start building what’s next! Start with $100 in AMD Developer Cloud credits and begin building right away. Get access to hands-on tutorials, workshops, and curated learning resources to help you build faster and go further. To unlock credits and join the hackathon, sign up for the AMD AI Developer Program. 🤝 Go solo or team up — ship something real. 💰 $21,500+ total prize pool & 1 × AMD Radeon™ AI PRO R9700 GPU up for grabs. 📍 On-site Venue (May 9-10): MindsDB SF AI Collective 3154 17th St, San Francisco, California, USA 📍On-site participation is by invitation only. Travel and accommodation expenses will not be covered. 🧑‍💻 Developers, builders, and innovators - bring your boldest ideas and let’s turn them into real AI systems on AMD.