Navigation and radar sensors: fusing data from several sources
Navigation and radar sensor makers need estimation and fusion algorithms and struggle to find specialists. How an outside review of a design helps.
Book a free 30-minute callControl and optimisation since 2007: CERN helium cooling · IAV wind turbines · SkySails airborne wind
Fusing inertial, satellite and laser-scanner data
What happens. A navigation unit has to combine inertial sensors, satellite positions and laser-scanner data into one position, and keep working when one source drops out. The algorithms must be designed, tested against recorded data and implemented on the device.
What it costs. Development relies on a few estimation specialists. Weak fusion shows up as drift, jumps in position when a source returns, and loss of accuracy in tunnels, canyons and indoor areas.
Why the usual fix fails. Libraries of standard filters give a start, but not the tuning of noise models and the checks of consistency that the product needs. Recruiting these specialists takes long.
How a fixed-price remote diagnosis checks it. From recorded sensor data and the sensor specifications I check how the current fusion behaves in the difficult cases and what a better estimator would change. The fixed-price remote diagnosis gives your team an independent view of the estimation design.
Background: Computing unmeasured states for control
Radar and camera fusion, tracking and nonlinear estimation
What happens. A radar maker combines radar detections with camera images to track objects. The motion of the objects and the measurements are nonlinear, and the tracks must be stable when objects are close, hidden or lost for a moment.
What it costs. Tracks that jump or split, objects that are lost at the wrong time, and tuning rounds that eat development months.
Why the usual fix fails. A standard tracking filter works in the simple case. Making it robust in dense traffic takes experience in nonlinear estimation that few teams have in-house.
How a fixed-price remote diagnosis checks it. From recorded radar and camera data with reference tracks I check where the tracker loses accuracy and which estimation method would help. The fixed-price remote diagnosis tells your team how much of the problem is in the filter and how much in the sensor data.
Background: Computing unmeasured states for control
More problems that concern this industry
These problems are described on the page of the industry where they matter most:
- Lightweight estimation with event cameras is hard to get small and robust (Space companies)
- Sensors on moving platforms must stay on target (Security and defence technology)
Check it on your plant first: a fixed-price remote diagnosis
EUR 4,500 net · 5 working days · remote · availability on request · freelance or through temporary agency work
Within a fixed scope and at a fixed price, I review your data, control structure and measurements remotely and tell you whether the problem on this page is behind your symptoms and what fixing it would take. You get a short summary for decision-makers and a technical report for your engineers.