Robotics subsystems · 4D radar · sensor fusion

Robotics subsystems built around radar we design ourselves

Most robotics shops integrate a radar someone else built. Here the radar front end is designed in-house — FMCW at 5.8 and 24 GHz, processed on FPGA — then fused with camera and inertial data and tied to motor drive. Perception, actuation, and the sensor itself, from one accountable engineer. 20+ years.

What sets this apart

A robotics subsystem is the perception-and-actuation layer of a robot — the sensors that see the world, the processing that interprets it, and the drives that move in response. The rare part here is that the radar itself is developed, not purchased: antenna arrays, RF front end, and FPGA-side signal processing are built to the robot's needs, not bolted on from a module datasheet.

That vertical reach matters where an off-the-shelf radar does not fit — a custom frequency plan, a tight enclosure, an unusual field of view, or a cost target a commercial module cannot meet. When the sensor and the software come from the same hand, calibration and fusion stop being an integration gamble.

Who needs it

Robotics teams whose perception fails where cameras and LiDAR struggle — fog, dust, darkness, glass, reflective surfaces — and who need radar to fill the gap.
Products that cannot absorb the cost or size of a commercial 4D radar module and need a front end designed to the target.
Mobile and industrial robots needing sensor fusion — radar with camera or inertial data — for reliable detection and velocity.
Systems that need perception tied cleanly to motor drive and positioning, not delivered as two disconnected black boxes.
Teams needing FPGA-side signal processing for latency a general-purpose processor cannot hold.

How the work runs

Sensing requirements

Range, resolution, field of view, update rate, and the environment the robot ships into define the radar plan and the fusion strategy — before any hardware is committed.

Radar front end

FMCW design at 5.8 or 24 GHz: link budget, patch antennas and arrays on FR-4, and an RF front end laid out for signal integrity. Calculated first, measured on real hardware second.

FPGA signal processing

Range-Doppler processing and detection on FPGA for deterministic, low-latency output the control loop can rely on.

Sensor fusion

Radar fused with camera or inertial input for robust perception — object separation and velocity that no single sensor delivers alone.

Actuation and integration

Perception tied to motor drive, positioning, and the robot's control system, then validated on real hardware under real operating conditions.

Why MPV-LAB for this

The radar is ours. RF front end, antennas, and FPGA processing are designed in-house — not integrated from a module. That is the rare combination in this field.
One hand from sensor to drive. Perception, fusion, and actuation come from the same engineer, so calibration and timing are engineered, not negotiated between vendors.
FPGA where latency matters. Signal processing on Artix-7 gives the deterministic, low-latency output a control loop needs.
Proven on real hardware. Radar boards designed, fabricated, and measured — not slideware. The 3D portfolio shows a robot drive unit taken to assembly.
You keep everything. Radar design, FPGA sources, and fusion code are handed over — no locked binaries, no lock-in.

Common questions

Do you design the radar or integrate an existing one?

Design it. The RF front end, antenna arrays, and FPGA-side processing are built to the robot's requirements. If a commercial module genuinely fits better, you get that recommendation honestly — but the option to build a front end no off-the-shelf part can match is the differentiator here.

What frequencies do you work at?

FMCW at 5.8 and 24 GHz, with 4D imaging radar and fusion of radar data with camera or inertial input. The frequency plan follows range, resolution, and regulatory constraints, decided against your target with numbers first.

Can you fuse radar with a camera or LiDAR we already use?

Yes. Sensor fusion combines radar velocity and all-weather robustness with the spatial detail of a camera or LiDAR, for perception that survives conditions where a single sensor fails.

Do you handle motor drive and positioning too?

Yes. Perception is tied to motor drive, positioning, and the control system, so the subsystem arrives as one coherent unit rather than disconnected sensing and actuation.

Who owns the design?

You do, once the milestone is paid. Radar design, FPGA sources, and fusion code are handed over in full. Any other engineer can pick the package up and continue.