Thrilled to announce a major milestone for our end-to-end driving stack at @teamminuszero — point-to-point autonomous navigation without HD maps using language based prompts! With our mapping interface, users can simply enter a destination, and our end-to-end models autonomously drive the full route to get there. Much like a human driver, our system interprets high-level navigation cues — such as “turn left” or “turn right” — from standard mapping services and plans its trajectory accordingly. Since we only rely on high-level guidance, there’s no dependency on dense HD maps, enabling the system to scale rapidly and generalize to new roads. This also means our stack can integrate across multiple mapping solutions with ease. All demo videos below were captured on roads never seen during training 👇 #foundationalmodels #autonomousdriving #fsd #llms
