Consultant / Expert Reviewer - UAV Sensor Fusion and Navigation Estimator Architecture in Bedford

We are looking for an experienced technical consultant to review and advise on the design of a sensor fusion and navigation estimation process for a UAV / drone autonomy project. We have already developed an initial design for the fusion architecture, including a local estimator, a global alignment layer, source-health gating, confidence handling, and staged implementation milestones. We are looking for an expert reviewer to assess whether the design is technically sound, operationally safe, and suitable for further prototyping, replay testing, and field validation. This is not intended as a full-time role at this stage. We are looking for focused expert review, practical challenge, and technical recommendations.

The proposed architecture separates fast, short-term movement tracking from slower position correction handling. The short-term estimator is intended to maintain a smooth and continuous understanding of the drone’s recent movement using onboard sensors such as the IMU, optical flow, range altitude, barometer, GPS velocity where available, and visual odometry when healthy. A separate correction layer handles slower updates from GPS position and external localisation sources to support mission and safety decisions during degraded navigation.

Scope of review

• Reviewing the proposed sensor fusion and navigation estimator architecture, including the split between short-term local movement tracking and slower global position correction;
• Assessing the proposed estimator design, including state choices, measurement models, update strategy, and whether an error-state EKF or similar approach is appropriate for the first implementation;
• Challenging the proposed design and identifying any areas that may be over-modelled, under-modelled, or likely to create false confidence;
• Reviewing how sensor input and external localisation corrections should be fused, trusted, rejected, or downweighted;
• Reviewing how intermittent external position corrections should be introduced without corrupting the live movement estimate or destabilising flight behaviour;
• Reviewing navigation confidence and lock-state modelling, including GPS lock, external correction lock, dead reckoning, degraded navigation, and uninitialised states;
• Reviewing reset/reanchor policy, software structure, ROS2/PX4 integration boundaries, and the division of responsibility between the companion computer, mission logic, safety logic, and flight controller.

Relevant experience

• Experience in sensor fusion, state estimation, or navigation systems for drones, robotics, autonomous vehicles, or mobile platforms;
• Knowledge of Kalman filtering, error-state EKFs, pose estimation, or similar estimator design approaches;
• Experience with GPS-degraded or GPS-denied navigation;
• Working with sensors such as IMUs, optical flow, rangefinders, barometers, GPS, visual odometry, VIO, or SLAM systems;
• Handling noisy, delayed, intermittent, or confidence-scored sensor inputs;
• Source-health gating, covariance handling, outlier rejection, and estimator consistency;
• Familiarity with ROS2, PX4, ArduPilot, MAVLink, or similar robotics and drone platforms within a Python programming environment.

Ideal profile

The ideal person is a practical technical expert who can quickly understand an existing estimator design, ask the right questions, identify weak assumptions, and explain which parts of the architecture are likely to work in practice and which may become technical risks. We are particularly interested in someone who can bridge theory and implementation: someone who understands the mathematics of state estimation, but can also reason about real sensors, noisy data, timing problems, coordinate frames, drone flight behaviour, estimator failure modes, and operational safety. The right person should be comfortable challenging the design constructively. We are not looking for a rubber-stamp review. We want someone who can help us make the system more robust, inspectable, and testable.

Form of engagement

• Remote;
• A small number of expert review sessions;
• Hourly consultation or a short advisory package;
• Possible follow-on work if both sides agree it is useful;
• Initial review may include architecture diagrams, design notes, ROS topic assumptions, estimator state definitions, logs, replay data, or code snippets.

In your response, please include brief information about:

• Your experience with UAV, robotics, or autonomous system sensor fusion;
• Your experience with state estimation, EKF/ESKF design, Kalman filtering, VIO, SLAM, optical flow, IMUs, GPS, or GPS-denied navigation;
• Whether you have worked with PX4, ArduPilot, ROS2, MAVLink, TF frames, or bag replay workflows;
• Examples of relevant systems, projects, research, or field deployments you have worked on;
• The tools, languages, and platforms you are most familiar with;
• Your availability and hourly rate.

The project operates in the area of UAV autonomy, degraded-GPS navigation, and onboard localisation. Further technical details can be shared after an initial conversation and, if required, an NDA.