Drone makers: flight control and tuning rest on few experts

Flight and vehicle test experts are scarce

What happens. Flight tests need people who can plan a test, read the logs and say what the vehicle is doing. In a small company there are one or two of them, and they are in every test.

What it costs. Delays: tests wait for the expert, evaluation waits for the next free day, and findings reach the design late.

Why the usual fix fails. Hiring is slow for such a profile. More tests without more evaluation capacity only fill the data folder.

How a fixed-price remote diagnosis checks it. From your flight logs, test plans and a description of the vehicle I check which tests could be replaced or prepared by simulation and what the logs already show. The fixed-price remote diagnosis tells you where simulation first and flight-data analysis can save test flights.

Position drifts in wind and without a satellite signal

What happens. A drone holds its position by combining several sensors. In gusty wind, or when the satellite signal is poor or gone, the position estimate drifts and the drone wanders.

What it costs. Failed missions, hard landings and lost vehicles. Operators fly with larger safety distances than the job needs.

Why the usual fix fails. A better single sensor only moves the problem. Fixed filter settings are tuned for one condition and fail in the next.

How a fixed-price remote diagnosis checks it. From flight logs of all sensors, including the cases with weak signal, I check where the estimate drifts and why. The fixed-price remote diagnosis tells you whether a better fusion of the sensors you carry can hold position, and what extra sensor would matter.

Background: Computing unmeasured states for control

The wind at the drop point is unknown

What happens. A load dropped from the air is carried by the wind on its way down. The wind at the drop point is not the wind on the ground, and it is rarely measured.

What it costs. A miss distance at the landing point that can make a delivery useless or put it out of reach.

Why the usual fix fails. Wind forecasts are coarse and old by the time of the drop. A fixed correction factor fits one day only.

How a fixed-price remote diagnosis checks it. From the flight data of previous drops and the wind data you have, I check whether the wind along the path can be estimated from the vehicle’s own motion. The fixed-price remote diagnosis tells you whether wind estimation could shrink the miss distance, and what it needs from the vehicle.

Background: Computing unmeasured states for control

Gimbal stabilisation needs identification, filters and fast code

What happens. A gimbal must keep a camera steady while the platform moves. The design needs a model of the mechanics, a filter for the sensors and code that runs fast on a small processor.

What it costs. Image blur and jitter, long tuning on the bench for each variant and a product that is good in the lab and weaker in the field.

Why the usual fix fails. Hand tuning one axis at a time ignores the coupling between axes, and copying settings between variants brings surprises.

How a fixed-price remote diagnosis checks it. From identification measurements on the gimbal and the specification I check which model and controller structure fit. The fixed-price remote diagnosis tells you what stabilisation performance is within reach and what it needs in code and hardware.

Background: Fast control on embedded hardware

Model-based flight control for large unmanned helicopters, tested in the loop

What happens. Large unmanned helicopters need flight control that is built on a model of the aircraft, with the software tested in a simulation loop before it flies.

What it costs. Each flight of a heavy aircraft is costly and risky, and problems found late are expensive. The team must have both the aircraft dynamics and the test setup under control.

Why the usual fix fails. Gains tuned in flight take many sorties. A simulation that does not match the helicopter gives false confidence.

How a fixed-price remote diagnosis checks it. From the aircraft data, existing flight logs and the simulation setup I check how well the model matches and where tests in the loop can replace flights. The fixed-price remote diagnosis tells you what a model-based flight control programme would need to start.

Background: Control of drones and aerial systems

Simulation, navigation and control are built separately for each flight phase

What happens. An unmanned aircraft passes through take-off, climb, cruise and landing. The reference simulation, the navigation and the guidance and control for each phase are made by different people and with different models.

What it costs. Integration work when the phases are joined, and surprises at the transitions between them.

Why the usual fix fails. Adding a layer of mode switches treats the symptom. A simulation made for one phase does not carry over to the next.

How a fixed-price remote diagnosis checks it. From the aircraft data, the phase designs and flight logs I check where models and interfaces differ between phases. The fixed-price remote diagnosis tells you whether one reference simulation across all phases is feasible, and how much of the existing work could be reused.

Background: Control of drones and aerial systems

Control laws for fast airframes must come from flight data

What happens. For a fast new airframe there is no textbook model. The control laws have to be built from identification on flight data and the data covers only some flight conditions.

What it costs. Slow progress from flight to flight, and a controller that is good in the tested range and unknown outside it.

Why the usual fix fails. Wind-tunnel data is expensive and incomplete. A model fitted by hand to the flights fits those flights.

How a fixed-price remote diagnosis checks it. From existing flight logs and test plans I check how well the dynamics can be identified and which manoeuvres would add the most information. The fixed-price remote diagnosis tells you what control laws the data can support and where more tests are needed.

Background: Control of drones and aerial systems

Flight tuning and flight-log analysis are done by hand

What happens. After each flight someone reads the logs, looks for oscillation or lag and adjusts gains by hand. It depends on a person’s eye and on time.

What it costs. Many flights for little progress, and tuning knowledge that sits in one head.

Why the usual fix fails. Auto-tune functions in the flight controller cover standard cases. A manual routine does not scale to a growing number of variants.

How a fixed-price remote diagnosis checks it. From a set of your flight logs I check how much of the tuning can be done by identification and analysis on the data. The fixed-price remote diagnosis tells you whether a routine from log to settings is possible, and which flights it would still need.

Background: Control of drones and aerial systems

A flight-control team is being built and has no lead

What happens. A drone maker is growing and needs a flight-control team, but nobody with the experience to lead it. The first controller was written by someone who has moved on.

What it costs. Slow development, rework on control code that nobody fully understands, and a product that depends on a few lines nobody dares to change.

Why the usual fix fails. Hiring a lead takes months in a thin market. A team without a lead tends to copy open-source settings and stay there.

How a fixed-price remote diagnosis checks it. From the current flight-control code, test records and the product plan I check where the risk lies and what a lead would have to take on first. The fixed-price remote diagnosis tells you what to fix now with outside help and what to prepare for a hire.

Inspection airships need flight autonomy

What happens. Airships for inspection, for example of power lines or gas pipelines, are slow and light, and wind pushes them around. Flying them autonomously along a line is harder than it looks.

What it costs. A pilot is needed for each flight, and the airship stays on the ground when the wind is not benign. The inspection cost stays high.

Why the usual fix fails. A standard autopilot from a drone is designed for fast aircraft and does not suit the large, slow dynamics of an airship.

How a fixed-price remote diagnosis checks it. From the airship data, wind records and the planned routes I check what the dynamics and the wind limit. The fixed-price remote diagnosis tells you whether autonomous flight along a line is feasible, and what sensors and control it would need.

Background: Control of drones and aerial systems