To Master Students: For interest in the proposed projects, contact us at: [email protected] 

Projects for Master Thesis - Open for academic period 2024-2025 

Isaac Lab: Aerodynamic Simulation of Lift-Generating Structures in Isaac Lab Using NVIDIA Warp
Supervisors: Kostas Alexis | Keywords: Isaac Labs
Available. 
This project, developed in close collaboration with the NVIDIA Isaac Lab team, aims to extend the Isaac Lab simulation framework with high-fidelity aerodynamic capabilities by integrating the XLB lattice Boltzmann fluid solver to simulate airflow around lift-generating structures. Currently, Isaac Lab lacks native support for realistic aerodynamic force modeling, which limits its applicability for fixed-wing UAV research and other aerospace applications. The proposed work involves developing a plugin that couples Isaac Lab with XLB, enabling bidirectional exchange of physical quantities: aerodynamic forces (lift, drag, and moments) computed by XLB are fed back into the Isaac Lab rigid body simulation, while updated system states such as position and orientation are passed to XLB to advance the fluid simulation.
Isaac Lab: Multicopter System Identification Tool for Isaac Lab Using Newton
Supervisors: Kostas Alexis | Keywords: Isaac Labs
Available. 
This project, developed in close collaboration with the NVIDIA Isaac Lab team, aims to bridge the sim-to-real gap for multirotor platforms by building a differentiable multicopter model in Newton and integrating it with Isaac Lab as a system identification tool. Accurately matching a simulation to a real physical vehicle is one of the core challenges in robotics and autonomous flight - small discrepancies in motor dynamics, inertial properties, or aerodynamic coefficients can cause controllers trained in simulation to perform poorly when deployed on real hardware. The proposed work centers on constructing a differentiable multirotor dynamics model within Newton, exposing all physically meaningful parameters - such as motor time constants, thrust and torque coefficients, inertia tensor components, and drag terms - as trainable quantities.​
Isaac Lab: High-Level Flight Stack Integration with PX4 in Isaac Lab
Supervisors: Kostas Alexis | Keywords: Isaac Labs
Available. 
This project, developed in close collaboration with the NVIDIA Isaac Lab team, aims to enable closed-loop simulation between Isaac Lab and PX4, one of the most widely used open-source autopilot flight stacks for autonomous aerial vehicles. While Isaac Lab provides a powerful environment for physics simulation and reinforcement learning, it currently lacks native integration with real-world flight controller software, creating a gap between simulation-trained policies and deployable autopilot systems. The proposed work involves developing a dedicated Isaac Lab plugin that runs the PX4 flight controller software directly alongside the Isaac Lab physics engine, establishing a real-time bidirectional communication interface for exchanging control commands and state estimates between the two systems.
Planning: Toward Efficient Language-Grounded Semantic Exploration
Supervisors: Kostas Alexis | Keywords: VLMs, Path Planning, Autonomy Stack
Available. 
Autonomous exploration in unknown environments is a key capability for autonomous robots. Existing exploration planners typically rely on geometric information, prioritizing areas that maximize the exploration of unseen space. In our current system, the OmniPlanner exploration framework guides aerial and ground robots by generating collision-free trajectories that maximize a geometric exploration gain derived from the environment geometry. While effective for exploration coverage, this purely geometry-based approach does not account for semantic relevance, which is important for tasks such as object search. 

This project aims to extend the current geometry-based exploration strategy by integrating language-grounded semantic information into the planning process. Specifically, we propose introducing a semantic gain component using a Vision Language Model (VLM) such as CLIP. Image observations collected by the robot will be encoded by the VLM into language-aligned features and projected into the robot’s map representation. These semantic features will then be transformed into a semantic gain, enabling the robot to prioritize the exploration or regions that are not only geometrically informative but also semantically meaningful with respect to a given task.
Learning: Safe Reinforcement Learning for Robot Navigation in Unstructured Environments
Supervisors: Kostas Alexis | Keywords: Robot Learning, RL, CBFs, Autonomy Stack
Available. 
Navigation in novel and unstructured environments imposes operational risks to autonomous robots, as tracking errors and odometry drift may accumulate and lead to collisions. To augment the classic map-plan-act paradigm, modern autonomy stacks may feature learned visuomotor control policies, which directly make use of the available sensory information to augment the standard control task with collision avoidance capabilities to attenuate the risk of failures and collisions. This project aims to investigate the use of certificate functions from classical control theory (namely Control Barrier Functions (CBFs)) in a reinforcement learning task to improve overall safety and sample efficiency. The goal is to investigate theoretically-informed methods for utilizing CBFs directly during the training of visuomotor control policies, that enforce constraint satisfaction for safe navigation.
Learning: One Control to Rule them All
Supervisors: Kostas Alexis (NTNU) | Keywords: Robot Learning, RL, Aerial Robots
Available. ​
Reinforcement learning is increasingly being utilized in robot control tasks. Conventionally, an RL policy is attuned to a particular vehicle by exploiting its particular dynamics in simulation-based learning. In this work we seek a fundamental shift in this paradigm: we aim to create a "single policy for the flight control of multirotors" such that one unified policy can control any arbitrary multirotor aerial robot. We are looking to investigate the correct definition of the observation space of the policy, an expressive neural architecture and an appropriate strategy for learning (including reward shaping) such that this unified control policy can be derived and reach at least on-par performance with the current state-of-the-art (but without vehicle-specific tuning requirement).​
Perception: Multi-modal SLAM for Fixed-wing UAVs
Supervisors: Kostas Alexis (NTNU) | Keywords: SLAM, Fusion
Available. ​
Simultaneous Localization And Mapping (SLAM) is now a mature technological field. Nevertheless the majority of the methods are tailored to low-speed aerial, ground or underwater robots. Limited work has been conducted focusing on high-speed fixed-wing Unmanned Aerial Vehicle (UAV) navigation especially when this is about flying very close to terrain or within clutter (e.g., inside a forest or within urban zones). This project aims to build upon ARL's Unified Autonomy Stack - and especially our Multi-modal SLAM system - in order to develop a Vision-LiDAR-Radar-IMU fusion system for resilient SLAM in high-speed fixed-wing flight. 
Planning: Exploration Path Planning for Nonholonomic Robots
Supervisors: Kostas Alexis (NTNU) | Keywords: Path Planning, Autonomy Stack
Available. ​
Exploration path planning has long history in the domain of robotics. The Autonomous Robots Lab has developed one of the most widely utilized methods and software solutions (GBPlanner aka OmniPlanner). While the solution operates across aerial, ground and underwater robots it does not appropriately support nonholonomic platforms such as fixed-wing UAVs. This work package aims to extend our solution (https://github.com/ntnu-arl/gbplanner_ros) such that it involves a different sampling-based planning kernel capable to support nonholonomic vehicles with a focus on fixed-wing UAVs and underwater AUVs. This is a great opportunity especially for those interested in supporting the open-source community. 
Perception: Very-Long-Range Vision-based Navigation for Aerial Robots
Supervisors: Kostas Alexis (NTNU) | Keywords: SLAM, Autonomy Stack
Available. ​
Conventional visual- and visual-inertial methods for odometry estimation on aerial robots perform well in short-range missions, typically below few kilometers. Simultaneously, they are mostly verified with the robot/camera observing the scene from close proximity. On the other hand conventional vision-based localization for larger systems - such as in the defense domain - relies on a priori known maps on which localization takes place. This work targets the development of a (a) robust and (b) high-performance Visual-Inertial Odometry solution for fixed-wing UAVs capable of flying missions in the range of 20-50km with minimal error. We aim to investigate the collective potential from increased camera sensors, high-performance yet increasingly more low-cost MEMS IMUs, time-synchronization and importantly high-quality visual registration especially through progress in deep learning. ​
Legged Robots: Autonomous Lunar Surface Mapping and Resource Detection for the LOOK 2026 Competition
​Supervisors: Kostas Alexis (NTNU) | Keywords: SLAM, Autonomy Stack
Available. ​
The exploration of the lunar surface presents significant challenges due to the lack of Global Navigation Satellite Systems, harsh lighting conditions, and complex terrain. The LOOK competition (Lunar rObOt Konkurranse), organized by the Norwegian Space Cluster, tasks student teams with building robots capable of mapping the lunar surface and identifying critical resources. This project aims to leverage the state-of-the-art ANYmal quadrupedal platform and the Unipilot autonomy module available at the Autonomous Robots Lab to create a competitive entry for the June 2027 competition. The research will focus on the integration of elevation mapping and real-time object detection to navigate and characterize a simulated lunar environment. By utilizing existing inspection cameras and advanced sensing suites, the student will develop a mission logic capable of autonomous exploration and resource localization. This work provides a unique opportunity to apply systems engineering and robotics expertise to a national competition with high visibility in the Norwegian space sector. ​
Legged Robots: Payload-Adaptive Walking, Jumping, and In-Flight Attitude Control for a Jumping Quadruped in Reduced Gravity
​Supervisors: Kostas Alexis (NTNU) | Keywords: SLAM, Autonomy Stack
Available. ​
For future robotic missions to planetary bodies, legged platforms must be capable of carrying diverse scientific instruments such as cameras, spectrometers, and LiDAR systems. However, the addition of a significant payload, which can reach up to 2.5 kg, fundamentally alters the robot’s mass distribution, center of gravity, and inertia tensor. These changes can lead to instability during walking or significant landing inaccuracies during jumping maneuvers if the control policies are not payload-aware. This project thesis aims to develop payload-adaptive reinforcement learning policies using the olympus_lab framework. The research will focus on creating a robust controller capable of maintaining high performance in walking, precise jumping, and in-flight attitude reorientation despite varying payload configurations. By utilizing techniques such as domain randomization and history-dependent policy architectures, the robot should learn to implicitly infer its physical properties from motion feedback. The results will be validated in a simulated Martian environment to ensure that the Olympus platform can safely transport scientific equipment across challenging terrain.
Legged Robots: Unified Omnidirectional Jumping and Walking for a Quadruped in Reduced Gravity using Reinforcement Learning
​Supervisors: Kostas Alexis (NTNU) | Keywords: SLAM, Autonomy Stack
Available. ​
As planetary exploration targets scientifically valuable but inaccessible regions such as Martian lava tubes, steep crater rims, and boulder-filled terrain, the need for agile robotic mobility becomes paramount. While the Olympus jumping quadruped has demonstrated high-performance jumping and robust walking, current control strategies developed for the Olympus quadruped rely on separate, task-specific policies for vertical and horizontal maneuvers. This project aims to bridge this gap by developing a unified reinforcement learning policy within the olympus_lab framework. The main objective is to enable the robot to seamlessly transition between walking and jumping by utilizing a 3D takeoff velocity vector as a primary control input. This approach will allow for arbitrary jump trajectories, providing the flexibility required for complex exploration scenarios where fixed jump profiles are insufficient. We also want to incorporate yaw rotation targets. The work will leverage existing curriculum-based learning strategies and physics-informed reward densification to achieve precise landing and robust traversal in simulated Martian gravity. If progress is sufficient, the developed policy will be validated on the Olympus hardware to demonstrate successful Sim2Real transfer in earth gravity.
Humanoids: Unified Autonomy Stack transition to Humanoid Robots
​Supervisors: Kostas Alexis (NTNU) | Keywords: SLAM, Autonomy Stack
Available. ​
The Autonomous Robots Lab has developed an autonomy stack that has been verified to operate across aerial, ground and partially underwater robots: https://ntnu-arl.github.io/unified_autonomy_stack/

This project aims to extend the application of this stack to humanoid robots, systems expected to dominate in the future of robotics. You will be provided a humanoid robot and the autonomy hardware reference module of the lab with the goal to enable the high-performance integration of the two targeting tasks such as autonomous navigation, exploration, area inspection, and object-level scene reasoning.
Learning: Unified Control for Floating-base Robots
​Supervisors: Kostas Alexis (NTNU) | Keywords: Robot Learning, RL
Available. ​
Floating-base robots – such as rotorcraft aerial vehicles and underwater ROVs – share a common characteristic: their motion is governed by fully actuated or underactuated dynamics in free space, without direct support contacts. Despite this shared structure, control strategies are typically developed in a robot-specific manner, limiting transferability across morphologies and environments.

This project aims to develop a unified control policy for floating-base robots, leveraging recent advances in reinforcement learning and foundation models. The goal is to design a generalist control policy capable of achieving high performance across different robot morphologies by incorporating explicit morphology conditioning. This approach aligns with the broader vision of a unified autonomy architecture capable of generalizing across embodiments while retaining high performance.

The project will explore how shared structure in dynamics and perception can be exploited to enable cross-platform generalization, targeting both aerial and underwater robotic systems.
Hardware: Extinguishing Wildfires with Guided Airdrops
Supervisors: Kostas Alexis, Tor Arne Johansen | Keywords: XPRIZE Wildfire Challenge
Available. ​
This project aims to develop a specialized payload to be ferried by a fixed-wing unmanned aerial vehicle (UAV) that can be used to suppress a wildfire detected at its very early stages after ignition. The envisioned payload shall be released by the ferrying UAV at high speeds and using control surfaces (w/o propulsion) shall guide itself accurately to drop on the desired point of the fire area. Upon impact, the payload shall release a specialized fire retardant or other technology to eliminate the early-detected small wildfire area using a limited number of such airdrops. ​​
 
Learning: World Models for Semantic Inspection
Supervisors: Kostas Alexis | Keywords: Embodied Autonomy
Available
Abstract: World models have emerged as a powerful paradigm for learning compact representations of an environment and predicting future states, enabling efficient decision-making in autonomous robotics. This project explores the application of world models to semantic inspection tasks, where a robot equipped with visual sensors must construct a semantic understanding of its environment while maximizing inspection efficiency. By leveraging learned latent space representations, predictive modeling, and reinforcement learning, the system aims to generate optimal inspection paths that balance exploration, occlusion handling, and semantic information gain. ​
Design: Transforming Aerial Robots
Supervisors: Kostas Alexis, Paolo De Petris | Keywords: Aerial Robotics
Available. ​
This project aims to develop a novel reconfigurable multirotor for extreme maneuverability. In particular, we seek to design and built a multirotor with each motor being on the tip of a 5-bar manipulator mechanism thus allowing seamless reconfiguration of its shape and control allocation. Alongside the design, we seek to derive an adaptive control policy to optimally maneuver such a robot along agile trajectories demonstrating superior maneuverability compared to conventional fixed-configuration aerial robots. ​
  • Details - this project is open ended to be co-designed with the student. Contact us. ​​
  • Project Relevance: SPEAR (European Commission Horizon Europe), RESSORT (European Commission Horizon Europe)
  • Contacts: Kostas Alexis ([email protected])
Design: Soft Aerial Robots for Hard Tasks
​Supervisors: Kostas Alexis, Etor Arza | Keywords: Embodied Autonomy
Available. ​
 This project aims to develop soft aerial robots in order to explore the advanced potential such systems can enable when it comes to navigating through highly-cluttered environments (e.g., dense forests). ​
  • Details - this project is open ended to be co-designed with the student. Contact us. 
  • Project Relevance: SPEAR (European Commission Horizon Europe), RESSORT (European Commission Horizon Europe)
  • Contacts: Kostas Alexis ([email protected])
Design: Micro Demining Robots
​Supervisors: Kostas Alexis | Keywords: Demining, Excavation
Available. ​
This project focuses on the development of a modular, remote-operated micro-excavation system designed to bridge the gap between subsurface detection and safe object identification in high-risk environments. By replacing manual, high-stakes probing with a precision robotic platform, the research aims to significantly reduce the physical risk to field technicians while increasing the speed of clearance operations. The project centers on the engineering of a lightweight, man-portable chassis equipped with a specialized end-effector capable of "surgical" soil removal - using techniques such as low-pressure air displacement or precision mechanical scraping. Key research areas include the integration of low-latency haptic feedback for remote operators, the optimization of power-to-weight ratios for deployment in remote terrains, and the creation of a robust control interface that maintains high situational awareness. Ultimately, this work provides a scalable, cost-effective technical solution to one of the most dangerous bottlenecks in humanitarian field recovery, focusing on the delicate transition from locating a subsurface anomaly to its complete visual exposure.
  • Details - this project is open ended to be co-designed with the student. Contact us. 
  • Project Relevance: Norwegian Centre for Embodied AI 
  • Contacts: Kostas Alexis ([email protected])
Everything is a Robot: Making an autonomous robot in the form of everyday house items
​Supervisors: Kostas Alexis | Keywords: perception, Semantics
Available. ​
 This project aims to investigate the potential of turning every day house items, such as a coffee table or a desk lamp to robots and identify the possible use space and potential of such novel products. By the end of the thesis the goal is to have designed such a prototype and investigate its applicability and relevance in the real-world. This is a particularly open-ended project calling for the student creativity.
  • Details - this project is open ended to be co-designed with the student. Contact us. ​
  • Project Relevance: Open-ended
  • Contacts: Kostas Alexis ([email protected])

Own ideas - high risk projects!

Do you have your own idea about a robotics project? Are you willing to discuss a high-risk project with the understanding that things might not always work? Contact me and schedule a meeting!