State Estimation Engineer (Navigation & Sensor Fusion) – UAV Autopilot Platform

Department: Guidance, Navigation & Control (GNC)

Reports to: GNC Lead / Head of Engineering

Experience: 4–8 years in state estimation, sensor fusion or navigation systems

Employment type: Full-time, on-site (lab and flight-test access required)

Works closely with: Firmware Architect, Control Engineers, Simulation Engineer, Flight-Test team

About the Role

We are looking for a State Estimation Engineer to own the navigation and attitude estimation stack of our autopilot. You will design the estimation algorithms, prove them in simulation, port the mathematics into efficient and numerically robust flight firmware, and validate them in real flight. You will also own sensor calibration and the estimator's behaviour when sensors fail.

The estimator is what the vehicle "believes" about its own attitude, velocity and position. Your work directly determines how safely and precisely the aircraft flies.

Key Responsibilities

- Sensor Fusion and Estimation Algorithms
- Estimator design: Design, implement and tune estimators for attitude, velocity, position, wind and sensor biases: EKF, Error-State / Multiplicative EKF, UKF and complementary filters as appropriate.
- Sensor fusion: Fuse IMU, magnetometer, barometer, GNSS (position/velocity/heading), airspeed, rangefinder/LiDAR and optical flow, and plan for vision/external navigation inputs.
- Time alignment: Handle sensor latency, delayed-measurement fusion (state buffering / time-horizon fusion), different sample rates and timestamp alignment with the firmware's data pipeline.
- Innovation monitoring: Implement innovation gating, consistency checks and covariance health monitoring.
- Porting Mathematics to Firmware
- Model to code: Take algorithms from MATLAB/Simulink or Python prototypes to production-quality embedded C/C++ that runs within the CPU and memory budget on ARM Cortex-M.
- Numerical robustness: Ensure stability in single-precision float: covariance symmetry and positive-definiteness, Joseph-form or UD/square-root updates, normalisation, guarding against divide-by-zero and NaN/Inf propagation.
- Performance: Optimise for real time: code generation of Jacobians (e.g. SymPy or MATLAB Symbolic Toolbox), sparsity exploitation, fixed-size matrix libraries, and FPU/DSP instructions.
- Integration: Work with the Firmware Architect on loop rates, scheduling, interfaces and timestamps so the estimator meets its timing and latency budget.
- Bit-matching: Build back-to-back tests that show the firmware output matc
- Quaternion and Attitude Mathematics
- Own the attitude representation and rotation mathematics: quaternions, rotation matrices (DCM),



Euler angles and their conversions, with clearly documented conventions (Hamilton vs JPL, body/NED/ENU frames, active vs passive rotations).
- Implement attitude propagation (quaternion kinematics, coning/sculling compensation where relevant) and error-state attitude formulations.
- Handle singularities (gimbal lock), quaternion normalisation and sign ambiguity, and provide a well-tested shared maths library for the whole team.
- IMU and Sensor Calibration
- IMU calibration: Design factory and field calibration procedures for accelerometer and gyroscope: bias, scale factor, misalignment/cross-axis, and temperature compensation.
- Magnetometer calibration: Hard-iron and soft-iron calibration (ellipsoid fitting), motor/current interference compensation and declination handling.
- Other sensors: Calibrate barometer (temperature drift), airspeed sensor offsets, board/sensor orientation and level-horizon trim.
- Tooling: Build calibration routines in firmware and in the ground station, with stored calibration parameters and quality metrics.
- Characterise sensor noise using Allan variance and use it to set the filter's process and measurement noise.
- Validation and Verification
- Validate estimators in simulation (SITL/HITL) with realistic sensor models, then on bench, ground vehicles and in flight against reference truth (RTK GNSS, motion capture or a high-grade INS).
- Build log-replay tools to re-run the estimator offline on recorded flight data for tuning and regression testing.
- Define performance metrics (attitude/heading/position error, convergence time, consistency via NEES/NIS) and acceptance criteria per airframe.
- Maintain automated regression tests for the estimator in the CI pipeline.
- Failure Modes and Robustness
- Perform FMEA for the navigation system: sensor dropout, stuck or noisy sensors, IMU clipping and vibration, magnetic interference, GNSS loss/jamming/spoofing and multipath, baro disturbances from prop-wash or ground effect.
- Implement fault detection, isolation and recovery: redundant-sensor voting/selection, multiple estimator instances (lane switching), estimator reset and re-initialisation logic.
- Define degraded-mode behaviour (e.g. dead-reckoning, attitude-only mode) and the signals the failsafe state machine uses to trigger safe responses.




- Support flight-incident investigations by analysing logs and identifying estimation-related root causes.

Key Deliverables (First 6–12 Months)
- Production EKF running on target hardware within the CPU budget, with documented conventions and interfaces.
- Shared quaternion / rotation maths library with unit tests.
- IMU, magnetometer and barometer calibration procedures (firmware + ground-station tooling).
- Log-replay and estimator validation toolchain with defined performance metrics.
- Failure-mode analysis document and imple

Required Qualifications

- Bachelor's or Master's degree (PhD a plus) in Aerospace, Electrical, Mechanical Engineering, Robotics or a related field.
- 4+ years of hands-on experience designing and implementing Kalman-filter based estimators for real systems.
- Strong command of linear algebra, probability, rigid-body kinematics, quaternions and rotation theory.
- Proficiency in C/C++ for embedded targets, and MATLAB and/or Python for prototyping and analysis.
- Experience with inertial sensors and their error models, and with sensor calibration.
- Experience taking an algorithm from simulation to real hardware and flight/field testing.

Preferred Qualifications

- Experience with PX4 EKF2, ArduPilot EKF3 or comparable autopilot navigation filters.
- Experience with GNSS/INS integration, RTK and GNSS integrity/anti-spoofing.
- Experience with visual-inertial odometry or optical-flow fusion.
- Familiarity with fixed-point or single-precision numerical optimisation on microcontrollers.
- Experience with SITL/HITL simulation and flight-log analysis tools.

Tools and Technologies

- Languages: C, C++, Python (NumPy, SciPy, SymPy), MATLAB/Simulink.
- Libraries: Eigen or embedded matrix libraries, CMSIS-DSP.
- Tools: Log-analysis and plotting tools, Allan-variance tooling, simulation environments (SITL/HITL), Git.
- Reference systems: RTK GNSS, motion capture, tactical-grade INS for ground truth.

What Success Looks Like

- Attitude and position accuracy meet defined targets across all supported airframes and flight envelopes.
- The estimator remains consistent and recovers gracefully from sensor faults without loss of control.
- Calibration is quick, repeatable and can be done by field operators.
- Every estimator change is validated by replay and simulation before flight.

Soft Skills

- Rigorous and methodical, with strong mathematical intuition.
- Able to explain estimation behaviour clearly to firmware, control and flight-test colleagues.
- Comfortable working on the flight line and digging through logs to find root causes.mented sensor-fault handling.

hes the reference model.

📌 State Estimation Engineer (Gurugram)
🏢 Flybots Tech Services
📍 Gurugram