Personal Portfolio

Pedro Lima

Aerospace Engineer

drag to orbit · scroll / pinch to zoom · tap a body for its dossier

Pedro LimaAerospace Engineer · GNC · Space Robotics

Aerospace Engineer · GNC · Space Robotics

I build machines that fly, orbit, and explore — working where physics meets software. My focus is guidance, navigation & control (GNC) and space robotics: modeling, designing, and validating the control and estimation systems behind them, from ground-effect vehicles to free-floating platforms.

I'm currently a Research Specialist with the Space Robotics Research Group (SpaceR) at the University of Luxembourg, where my research covers classical control, robot learning and reinforcement learning for autonomous space systems. The aim: push what autonomous robots can do beyond Earth.

Works & Credentials

Publications, honors, and courses

Published work, honors, and courses that round out the engineering toolkit.

Publications

  • Fixed-Wing Robots in Ground Effect: A High-Fidelity Aerodynamic Simulation Framework for Flight Control — Journal · Under review · 2026
  • PINGU: Design, Characterization, and Validation of a Multi-Actuator Air-Bearing Spacecraft Emulator — Journal · Under review · 2026
  • Autonomous Spacecraft Docking and Fuel Transfer-Ready Alignment for Sustained On-Orbit Operations via Reinforcement Learning with Constraints — Conference · ESA GNC & ICATT · 2026
  • Can Learned Models Enhance Spacecraft Docking and Refueling in a Self-Sustaining Space Economy? — Conference · iSpaRo · 2026

Honors

Courses

  • Python for Computer Vision — OpenCV & Deep Learning — Course · 2023
  • ROS2 C++ Robotics Developer — Course · 2024

Education

Aerospace & space engineering — Lisbon and Rome

Aerospace and space engineering — from the fundamentals of aerodynamics, structures, and propulsion to guidance, navigation & control.

My master's thesis built an air-bearing testbed for spacecraft attitude control: an active mass-balancing system with three sliding masses that cancels residual gravitational torque, recreating a near frictionless, torque-free space-like environment. A PID controller and an extended Kalman filter (EKF) estimate and correct the platform's centre of mass, validated in both simulation and experiment.

Skills

Control, robotics, and the tools behind them

From control theory to CAD to code — the toolkit behind the work.

GNC & Robotics

  • Domains: Simulation
  • Domains: Control
  • Domains: Navigation
  • Domains: Guidance
  • Domains: Manipulation
  • Domains: Perception
  • Control & Estimation: LQR
  • Control & Estimation: MPC
  • Control & Estimation: MPPI
  • Control & Estimation: PID
  • Control & Estimation: EKF
  • Control & Estimation: UKF
  • Learning & RL: PPO
  • Learning & RL: SAC
  • Learning & RL: SHAC

Software & Tools

  • Programming: Python
  • Programming: C++
  • Programming: C
  • Programming: CUDA
  • Programming: MATLAB/Simulink
  • Programming: Julia
  • Middleware & Simulation: ROS2
  • Middleware & Simulation: NVIDIA Warp
  • Middleware & Simulation: NVIDIA Isaac Lab
  • Middleware & Simulation: NVIDIA Newton
  • Middleware & Simulation: OpenCV
  • Middleware & Simulation: ArduPilot/PX4
  • Tooling: Linux
  • Tooling: Git
  • Tooling: Docker
  • Tooling: LaTeX
  • CAD & Mechanical: SolidWorks
  • CAD & Mechanical: Onshape
  • CAD & Mechanical: Ansys

Languages

  • English — Fluent
  • Portuguese — Native
  • Italian — B1

Experience

Each moon is a role

Roles across space robotics, GNC, and control. Click a moon for the details of each position.

Research Specialist

Space Robotics Research Group (SpaceR), University of Luxembourg · Luxembourg

Research on autonomous robotic systems, where GPU-accelerated simulation is a common thread. Simulating thousands of environments in parallel is what makes modern control tractable: sampling-based methods like MPPI need the throughput, and learning methods need either that volume of experience or a differentiable model to backpropagate through. I build those simulators and the controllers that run on them.

The applications span free-floating spacecraft for on-orbit servicing, wing-in-ground vehicles, and more recently, robotic manipulation. Whatever the platform, the goal is the same: policies that survive the move from simulation to hardware to space.

  • Simulation: GPU-accelerated and differentiable physics (NVIDIA Warp, Isaac Lab)
  • Classical control: LQR, MPC, MPPI
  • Robot learning & RL: PPO, SAC, SHAC
  • On-orbit servicing: autonomous docking and fuel-transfer-ready alignment on a free-floating air-bearing platform
  • WIG vehicles: high-fidelity ground-effect aerodynamics for flight control
  • Flight testing: model-scale flights validating control systems against simulation
  • Robotic manipulation: an emerging strand of the work

Aerospace Engineer

Trisolaris Advanced Technologies · Lisbon, Portugal

Responsible for GNC and systems engineering on two EU-funded wing-in-ground (WIG) vehicle programmes — AIRSHIP, a fully electric transport vehicle for inter-island routes, and SEAWINGS, a surveillance UAV for defence (see Projects). Both fly within a wingspan of the water to exploit the ground effect, trading altitude for speed and energy efficiency.

  • Preliminary design, flight mechanics modelling & stability analysis
  • Vehicle dynamics modelling and simulation in MATLAB/Simulink and Python
  • Control & estimation: LQR, MPC; EKF/UKF
  • Systems engineering: full-scale prototype integration
  • Secure comms & mission management (PX4, ROS2, MAVLink, QGroundControl)

Projects

Each moon is a project — one has its own spacecraft

Selected work where physics meets software. Click a moon to open a project.

AIRSHIP

Autonomous, zero-emission wing-in-ground vehicle

A Horizon Europe project developing a new class of fully electric unmanned WIG vehicle (UWV) for inter-island and inland-waters transport. Flying within a wingspan of the surface lets the vehicle ride the ground effect — the cushion of high-pressure air trapped between wing and water — which buys it the speed of an aircraft at a fraction of the energy, with no direct emissions and far less noise.

I have worked on AIRSHIP across both consortium partners: at Trisolaris on integration and systems engineering for the vehicle itself, and now at SpaceR on control systems and simulation — a high-fidelity aerodynamic framework for flight control, built to capture ground effect faithfully enough that controllers tuned in it transfer to the real vehicle.

  • Developed ground-effect flight simulators from scratch: one in MATLAB/Simulink, one in Python with GPU acceleration via NVIDIA Warp
  • Simulations underpin control system design, RL training and validation
  • Control & estimation: LQR, MPC, RL-based control, EKF/UKF
  • Systems engineering: integrating structure, propulsion, electronics, sensors & software on the full-scale prototype
  • Model-scale flight tests validating the control systems against simulation

SEAWINGS

Wing-in-ground surveillance UAV for the sea/air interface

A European Defence Fund project developing a new class of autonomous surveillance drone operating at the sea/air interface. Skimming the surface in ground effect gives the vehicle long range, high payload and a low radar signature at low cost — a combination that suits intelligence, surveillance & reconnaissance (ISR), search and rescue, and logistics missions.

I was responsible for secure communications and the mission-management interface — the link between operator and vehicle, and the assurance that nobody else could use it.

  • Secure comms: MAVLink + PX4 + ROS2, with Gazebo SITL for encryption testing
  • Mission interface: custom QGroundControl (Qt) integrating secure MAVLink

Air-Bearing CubeSat Attitude Testbed

MSc THESIS · 2021–2022 · Grade 19/20

Testing a satellite's attitude control on the ground is hard: gravity dominates, and any offset between the platform's centre of rotation and its centre of mass produces a torque that swamps the effects you're trying to measure. An air-bearing platform removes friction, but not that residual torque.

This thesis built a testbed that cancels it. Three sliding masses actively shift the centre of mass onto the centre of rotation, recreating a near-frictionless, torque-free environment in which spacecraft attitude control can be tested realistically. A PID controller drives the masses, and an extended Kalman filter estimates the offset online and feeds it back.

  • Active mass balancing with three sliding masses
  • PID controller and an EKF with feedback of the estimated parameter
  • Raspberry Pi + Arduino avionics
  • Validated in both simulation and hardware experiments

Floating Platform

Three-DoF air-bearing spacecraft emulator

A pneumatic free-floating platform that rides frictionlessly on air bearings across the floor of SpaceR's Zero-G Lab, reproducing in three degrees of freedom the planar dynamics a spacecraft experiences in orbit. Eight nozzles and a reaction wheel stand in for a satellite's actuators, and a motion capture system tracks every run as ground truth.

It's where on-orbit servicing research meets hardware: policies trained in simulation have to survive real actuators, sensor noise and delays. I work on the platform and the autonomy that runs on it — docking and fuel-transfer-ready alignment — a natural fit for constrained RL and sampling-based predictive control.

  • Developed a high-performance, GPU-accelerated spacecraft simulator using NVIDIA WARP for sim-to-real transfer, enabling parallel multi-environment simulations with precise physics.
  • Eight-nozzle, three-DoF air-bearing spacecraft emulator: design, characterization & validation
  • Predictive control: LQR and MPPI over GPU-accelerated dynamics
  • Reinforcement learning for autonomous docking and alignment (PPO, SAC, SHAC)

© 2026 Pedro Lima