Gauthier Lechevalier has built a working prototype of a radar system that can identify what objects are made of — a capability that pushes past conventional radar, which typically measures only distance and velocity.

The key technical accomplishment, documented at gauthier-lechevalier.com, combines beamforming with signal-processing software capable of extracting material information from radar reflections. Beamforming is a technique that uses an array of antennas to steer and focus a radar beam in a specific direction, concentrating the energy on a target while suppressing noise from surrounding reflections. Standard radar tells you where something is and how fast it's moving. To identify what an object is made of from the echo alone requires detecting subtler signatures: how the radar wave interacts with the material's electrical properties, surface texture, and how it absorbs energy at different frequencies. These differences show up in the strength, phase shift, and polarization of the return signal.

Beamforming helps by creating a narrow, controllable beam that concentrates interrogating energy on a specific volume and filters out clutter from adjacent reflectors, raising the signal-to-noise ratio enough that the classification algorithms downstream receive cleaner data. The harder problem — converting that cleaned-up return into a material identification — lies in the software algorithms rather than the antenna design itself.

This kind of sensing sits between conventional microwave radar and dedicated millimetre-wave or terahertz spectroscopy. Full spectroscopic identification typically requires either very short wavelengths, expensive hardware, or both. Radar-based approaches trade some precision for cost, range, and the ability to see through packaging or clothing — which makes them practical for industrial inspection, security screening, and IoT sensing applications where dedicated spectrometry equipment is too expensive or bulky.

Lechevalier's work falls within a research tradition that has gained commercial momentum over the past decade as millimetre-wave radar chips in the 60 GHz and 77 GHz bands became affordable enough for independent developers to experiment with. Cheaper silicon shifted the engineering focus from building custom antenna hardware to designing the software that processes the signals. That software-first approach is visible in Lechevalier's design: the beamforming logic and material-classification algorithms are where the real engineering work sits.

The gap between a functioning prototype and a system ready for real-world deployment is worth noting. Material-sensing radar works reasonably well in controlled settings — a factory with stable conditions and known target materials — but generalizing across different geometries, dielectric properties, and environmental noise is harder. Published prototypes in this field often carry hidden assumptions about target distance, the angle of incidence, and the granularity of what you're trying to distinguish (metal versus non-metal is simpler than telling polymer grades apart). Lechevalier's documentation does not specify the operational range or classification accuracy across material types, so those performance questions remain open.

What the prototype establishes is a working demonstration that beamforming-controlled illumination can work alongside material-discriminating signal analysis using a form factor and cost accessible to independent engineers. That is not a trivial result. The combination of RF hardware and algorithms suggests sensing capabilities that could appear in embedded systems anywhere persistent, non-contact material monitoring would be useful — production-line quality control, waste sorting, or monitoring structural health in buildings and bridges.

Radar sensing more broadly has been expanding its capabilities in exactly this direction. Doppler radar became gesture recognition; range-finding became presence detection; now material discrimination is moving from academic research into prototyped hardware built by individual engineers using commercial RF chips. The pattern has been consistent: specialized and expensive, then accessible, then commercial. Where Lechevalier's work lands in that progression will depend on how the classification performs as conditions become less controlled — but the prototype places it solidly in the accessible phase.