Spanish research center Ikerlan is developing BEGI, a neuromorphic 4D sensor that mimics biological vision to capture depth and motion using a single lens and without frame-based capture. Built on a dynamic vision sensor (Prophesee IMX636, 0.9 megapixel) overlaid with a microlens array, BEGI operates using an event-driven model: each pixel fires only when light changes, generating sparse, low-latency data streams. The system uses feedforward stereopsis to calculate 3D depth and motion without a second camera, requiring minimal power — a critical advantage for spacecraft and edge robots where thermal budgets are tight.
Current limitations: the prototype achieves ~0.2 effective-pixel resolution due to hardware constraints, limiting range to under 1 meter and field of view to 20–40 degrees. Ikerlan is targeting near-term applications in industrial robot guidance (factory arms at 1-meter reach) and aerospace satellite docking, where short-range and high-precision sensing is the bottleneck. To scale BEGI to broader physical AI use cases (drones, autonomous driving, humanoids), the team must design a custom ASIC with 4K+ resolution and >10 meters range — a capital-intensive pivot that depends on industrial partnership and/or sovereign funding.
For architects: neuromorphic vision (event-based, pixel-level) is a decade-old academic topic but now materializing commercially as silicon. BEGI's limitation (range) is also its niche: aerospace and precision robotics reward power efficiency over field of view. If the ASIC succeeds, expect a category bifurcation: neuromorphic sensors for latency/power-critical tasks, frame-based RGB/depth cameras for general vision. The race is on between camera-industry heavyweights (Sony, OmniVision) and pure-play neuro-chip startups (Prophesee, OmniVision's Gemini) to own this segment before robotics OEMs bake in custom silicon.