Skip to content

Sky Robotics Research Program

The goal of the SkyRobotics program is to develop and validate methods for cooperative interaction between airborne (UAV) and ground-based (UGV) robots in dynamic, uncertain environments. The research focuses on physically guided motion planning, federated learning of world models, and real-time diffusion-based task allocation. The project combines state-of-the-art approaches from machine learning, mobile robotics and multi-agent systems.

Key research challenges

  • Sensory complementarity: the UGV has accurate short-range sensors, the UAV has a wide field of view with less accuracy.
  • Dynamic differences: UAV moves in 3D space and is susceptible to weather conditions, UGV - limited to terrain.
  • Communication asymmetry: the UAV can act as a communication hub for a group of UGVs.
  • Different energy models: the UAV's mission time is usually shorter than the UGV's, forcing dynamic energy and task management.


Proposed publication directions

  1. Physically guided trajectory generation for heterogeneous UAV-UGV pairs
    • Generative diffusion models with dynamics ties for both platforms.
    • Joint trajectory planning with physical requirements and mission objectives.
  2. Federation learning of world models in heterogeneous swarms
    • Teaching environment models without exchanging raw data.
    • Aggregation of physically interpretable local models (wind, friction) into a global model.
  3. Real-time diffusion task scheduling with energy consideration
    • Adaptive assignment of tasks based on power consumption and battery status.
    • Generative modeling of task assignment probability distributions.
  4. Formal verification of the safety of UAV-UGV cooperation
    • Reachability analysis for hybrid systems.
    • Guarantee of no collision at the mathematical level.

The SkyRobotics team is currently developing a UAV research platform and a mutual localization system for robot swarms.
The project is open to collaboration in the areas of modeling, simulation, federated learning and integration of physical robotic systems.

We look forward to working with you!

Sky Robotics Research Program Manager
Matthew Andrzejewski
mand@pjwstk.edu.pl