Live Perception & Spatial Navigation
Alex Kendall / Wayve Core Thesis: Traditional HD-map autonomous systems struggle with novel UK road features (narrow unmapped lanes, jaywalkers, rain blur) because rule-based hand coding cannot cover millions of long-tail edge cases. Embodied AI uses end-to-end computer vision and learned physics models to generalise safely across unmapped environments.
Risk Metrics & Safety Benchmark
Disengagements / 100mi
0.12
Target: < 0.05
Near-Miss Events
1
Edge Case Trajectories
Safety Score
94.2
100 = Parity Target
Human Redundancy
7.5 yrs
Projected Timeline
Human Driver Redundancy Timeline Curve
Year 2026 - 2038