Imagine a car that can 'see' perfectly through thick fog or rain without draining its battery. This project builds a brain-like computer chip that processes light signals instantly, similar to how human neurons work. It makes laser-scanning sensors much smaller, cheaper, and far more energy-efficient.
Current LiDAR systems are too power-hungry, expensive to manufacture, and struggle with accuracy in bad weather, limiting their use in mass-market autonomous systems.
A photonic reservoir computing chip and a Conceptor Control Loop (CCL) for robust signal processing.
If you are an autonomous vehicle manufacturer dealing with high power drain from sensors—this project developed neuromorphic CIM and RC architectures that reduce power consumption in object recognition tasks by 50%. This allows for safer navigation in adverse weather with less battery impact.
If you are a smart factory provider dealing with bulky, expensive 3D scanning hardware—this project developed integrated FMCW LiDAR that reduces packaging size and manufacturing cost. This enables high-resolution predictive maintenance in tight industrial spaces.
If you are a city operator dealing with the high cost of monitoring urban traffic and security—this project developed ultra-low latency signal processing for LiDAR. This improves the accuracy of real-time object detection across city grids while lowering energy costs.
The project targets a 50% reduction in power consumption for object recognition tasks and aims to lower overall manufacturing costs compared to current state-of-the-art LiDAR systems.
Based on available project data, the use of photonic integrated circuits based on reservoir computing principles is specifically intended to ease the manufacturing process.
Based on available project data, the project is currently in the research and development phase (2024-2027), and specific licensing terms are not yet disclosed.
The system uses FeFET-based Compute-in-Memory accelerators designed to support hybrid models of both Spiking Neural Networks (SNN) and Artificial Neural Networks (ANN).
The project period runs from October 1, 2024, to September 30, 2027, suggesting a proof-of-concept will be available by late 2027.
The consortium is heavily research-driven, consisting of 7 partners across 5 countries (BE, ES, FR, KR, NL). With 4 research institutes and 2 universities, the academic weight is high, while industrial representation is low at 14% (1 company). The coordination by IMEC (NL) suggests a strong focus on semiconductor fabrication and chip-level integration.
Contact Stichting IMEC Nederland for technical specifications on FeFET-CIM accelerators.
Contact us to identify licensing opportunities for neuromorphic LiDAR hardware.
