Overcoming the pencil-beam pointing problem without wasting frames on exhaustive sector sweeps.
Recurrent neural networks correlate vehicle steering angles and suspension roll with peer trajectories to predict beam arrival angles to 0.5 degrees.
When a truck blocks a direct line-of-sight beam, the AI instantly switches to a pre-computed roadside building reflection path within 200 microseconds.