MH

MECHATRONICS · ROBOTICS · CONTROL

Marius Halaas

MSc student and Research Assistant at the University of Agder, working with robotics, control systems, simulation and reinforcement learning.

View projects
Current focus

Reinforcement learning & path planning

Developing PPO-based control for coordinated rotor and DC-motor actuation, with the long-term goal of a physically validated digital twin.

Isaac LabPyTorchPPORobotics

SELECTED WORK

Engineering projects

Projects combining physical prototyping, modelling, control and learning-based methods.

VIDEO

Add assets/videos/hybrid-robot.mp4

Bachelor project · Physical prototype · 2026

Hybrid ground–aerial robot

Designed, built and programmed a hybrid robot prototype as part of a two-person bachelor project. The platform combines differential-drive ground locomotion with quadrotor propulsion in a single experimental system.

  • Mechanical design, CAD and physical prototyping
  • Onboard electronics, sensors and system integration
  • Integration of wheeled drive and quadrotor propulsion
MechatronicsSolidWorksEmbedded systemsControl
VIDEO

Add assets/videos/simscape-model.mp4

Research Assistant · University of Agder · 2026

Integrated modelling of a hybrid robotic system

Built a complete MATLAB/Simulink and Simscape model integrating quadrotor dynamics, differential-drive locomotion, rotor forces and ground contact.

  • DC-motor control for differential drive
  • Cascaded PID quadrotor control with rotor mixer
  • Rotor-assisted turning and incline-control studies
MATLABSimulinkSimscapePID

HYBRID CONTROL STUDIES

Rotor-assisted mobility

Two simulation studies using the integrated model to investigate how rotor thrust can extend the robot's ground-mobility capabilities.

VIDEO

Add assets/videos/thrust-assisted-turning.mp4

Simulation study · Turning

Thrust-assisted turning

Rotor thrust is applied during ground turning to assist the manoeuvre and study its effect on turning performance and achievable turning radius.

VIDEO

Add assets/videos/slope-control.mp4

Simulation study · Slope control

Rotor-assisted slope control

Two distance sensors mounted a known distance apart measure the distance from the robot to the slope independently. The difference between the two measurements is used together with the known sensor spacing to estimate the robot's angle relative to the slope. This relative-angle estimate is fed back to the controller, which regulates rotor assistance and vehicle attitude to maintain or increase useful wheel loading and improve traction on steep inclines.

VIDEO

Add assets/videos/rl-path-planning.mp4

Research Assistant · Ongoing

Reinforcement learning & path planning

Developing PPO-based policies for coordinated rotor and DC-motor control in a simulated hybrid robotic platform.

  • Rewards based on path deviation, energy use and traversal time
  • NVIDIA Isaac Lab / Isaac Sim simulation workflow with PyTorch
  • LiDAR SLAM, vision integration and physical validation planned
Isaac LabPyTorchPPOPath planning
VIDEO

Add assets/videos/bldc-imu-rig.mp4

Physical control project

IMU feedback control of a BLDC test rig

Built a closed-loop angle-control system on a physical BLDC test rig using an ESP32, IMU feedback, Kalman-filtered angle estimation and PID control. The controller supports multiple angle setpoints and continuously adjusts motor output based on the measured angular error.

  • Closed feedback loop using IMU-based angle measurements
  • ESP32-based controller with ESC commands sent using the DShot protocol
  • Kalman filtering for angle estimation and noise reduction
  • Tracking tests across multiple commanded angle setpoints
ESP32BLDCIMUKalman filterPIDDShot
Robotics coursework

Industrial robot programming

Programmed a Universal Robots manipulator and generated paths and trajectories from a scanned image.

Universal RobotsROSTrajectory generation
Mechanical development

Cycloidal gearboxes

Designed and built cycloidal reduction gearboxes for a robotic-arm development project.

Mechanical designCADPrototyping

ABOUT

Engineering across hardware and software

I am a Mechatronics MSc student at the University of Agder with a particular interest in robotics and control systems. My work spans physical prototyping, embedded systems, dynamic modelling, classical control and reinforcement learning.

I enjoy projects where mechanical, electrical and software design have to work together as one system, and where simulation can be connected back to real hardware.