Guidance · Navigation · Control

Carlos Gonzales

Aerospace engineer focused on GNC after four years building flight hardware and test equipment at Lockheed Martin.

Now developing spacecraft simulation and autonomous perception systems. Current Top Secret clearance. U.S. citizen.

Scroll ↓

Featured work · Guidance, navigation and control

Orbital docking simulator

Crew Dragon capsule approaching the station with mission telemetry in the docking simulator.
Play opens the docking scene shown in the background. The autopilot demo opens separately.

I developed a MATLAB/Simulink spacecraft dynamics model and an interactive docking simulator, comparing their responses during normal operation and thruster failures. The simulator connects six degrees of freedom, navigation and feedback control to a mission you can watch or fly.

Verified results

Offline deterministic plant verification, not real spacecraft validation or a live MATLAB link.

Explore the GNC systems
  • MATLAB / Simulink Executable dynamics models with 100 Hz integration, quaternion attitude and a 16-thruster reaction control system. Compared MATLAB, Simulink and browser simulator trajectories in nominal and stuck-thruster cases; designed an LQR controller and assessed closed loop stability.
  • Dynamics Clohessy–Wiltshire relative orbital motion and rigid body attitude dynamics with thruster forces, torques, fuel use and injectable faults. Verified against analytical solutions, coordinate transformation checks and conservation laws.
  • Navigation ANEES-gated 6-state EKF for relative translation, plus an attitude MEKF with gyro-bias estimation and star-tracker updates.
  • Docking autopilot V-bar approach guidance with PID and LQR feedback, plus constrained model predictive control for terminal approach. The autopilot regulates closing speed and alignment while enforcing approach corridor and capture limits.
  • Allocation & Safety Bounded force and torque allocation across the thrusters, with corridor monitoring and automatic abort burns when an approach becomes unsafe.
  • Verification Seeded Monte Carlo analysis scores docking outcome, propellant use and time margin across randomized dispersions. Regression tests cover dynamics, navigation and actuator behavior.

Source code and documentation →

Explore the architecture

Inside the simulator

How the simulation, flight software, visuals and verification tools connect.
  • Done
  • In progress
  • TBC / Not started

Solid arrows: existing paths. Dashed arrows: unfinished integrations.

Architecture map of 14 connected systems, from the React app and simulation core to telemetry, rendering, GNC tools and MATLAB. Green blocks are done, yellow are in progress, and red are not started.

Status applies to the scope of each block. The current renderer and offline MATLAB model are complete; Volumetric Weather is in progress. Live MATLAB / SIL / HWIL integration and remote Runpod execution remain planned.

Simulator architecture

Full simulator architecture with all 14 blocks and 20 connections.

September 16, 2026 snapshot. Use 100% zoom to read details, then scroll to explore.

Featured work · Autonomous driving · CMPE 249

3D perception from a single camera

Five FCOS3D prediction boxes projected onto a nuScenes camera frame.
FCOS3D predictions projected onto a nuScenes camera frame. Recorded pipeline output; benchmark results below.
75 ms / frame
Mean latency including image decode on RTX 4090, FP32, using nuScenes mini.
33% drop
Relative mAP loss at 0.4× brightness on nuScenes mini validation. Measured in the paper ↗

I built a ROS 2 pipeline that detects cars, pedestrians and other objects in 3D using a pretrained FCOS3D neural network and public nuScenes driving data. The work focused on getting the model’s outputs into the right coordinate frame, measuring performance and understanding where perception breaks down.

What I learned

Correcting output decoding and camera calibration made the performance comparisons useful. The brightness sweep showed where single-camera detections become unreliable; regression tests now cover the fixes.

Read the technical paper →

Source code and evaluation results →

Experience

Lockheed Martin

Sunnyvale, CA · Flight hardware

Electromechanical Engineer

May 2024 — Aug 2026

Drove flight hardware mechanical design, qualification test fixture development, product definition, and drawing release. Flowed top-level requirements from Systems, RF, manufacturing, and quality into mechanical and electrical documentation.

Mechanical Engineer

Jun 2023 — May 2024

Maintained flight and GSE drawing sets through Windchill change tasks. Owned BOM documentation, vendor coordination, and DFM feedback; tracked long-lead items against build schedules.

Engineering Aide

Jul 2022 — Jun 2023

Incorporated engineer redlines and GD&T corrections into drawings and processed change requests through engineering release.

Earlier work

Technical skills

Education

Credits

Earth surface, night-lights, and specular textures — NASA imagery (public domain), via the three.js example assets; terrain relief from NOAA ETOPO 2022 and USGS 3DEP elevation data (public domain).

Crew Dragon — “SpaceX - Dragon 2” by KUBAHA. Source · licensed under CC BY 4.0. Modified: scale normalized, nose cover opened, surface materials adjusted for the simulator lighting.

F/A-18C — “McDonnell Douglas F/A-18C Hornet” by Rhine_Lab_Muelsyse. Source · licensed under CC BY 4.0. Modified: rescaled to 17.06 m, deployed gear and stores hidden.

Simulator built with three.js and React Three Fiber. Open-source dependencies are MIT licensed.