Andrea Del Prete

Andrea Del Prete is an Associate Professor in the Industrial Engineering Department at the University of Trento (Italy) since 2022. His research focuses on robot control, reinforcement learning, trajectory optimization, and numerical algorithms for dynamic systems. He leads the Interdepartmental Robotics Lab (IDRA) and has previously held roles as a tenure-track assistant professor at the University of Trento (2019-2021), a research scientist at the Max-Planck Institute for Intelligent Systems (2018), and an associated researcher at LAAS-CNRS (2014-2017) working with the HRP-2 humanoid robot. Earlier, he conducted PhD and post-doc research at the Italian Institute of Technology (2010-2013) on iCub robot control.

• PhD in Robotics (2013) - Italian Institute of Technology
• MEng in Computer Engineering (2009) - University of Bologna
• BSc in Computer Engineering (2006) - University of Bologna

Dr. Del Prete specializes in merging learning and model-based techniques for safe robot control, particularly in legged systems. His work bridges trajectory optimization (TO) with reinforcement learning (RL) to overcome local minima challenges (CACTO/CACTO-SL algorithms) and develops robust controllers for humanoid and quadrupedal robots in unstructured environments. He explores viability kernels in MPC, safety certificates, and bi-level optimization for co-designing hardware/control policies. Key application areas include mountain rescue robotics (ALPINE platform), aerial maneuver recovery, and energy-efficient legged locomotion.

His recent publications (2023-2025) emphasize numerical optimization algorithms, multi-contact locomotion, and hybrid control frameworks. Topics span from analytical integral optimization (2025) to climbing robots for mountain operations (2025), demonstrating a trajectory from theoretical algorithm development to real-world robotic applications. Research keywords include robotics, numerical optimization, and machine learning, with sub-fields like MPC for dynamic systems, humanoid control, and terrain adaptation.

As an educator, he teaches advanced courses on:

• Optimization and Learning for Robot Control (48-hour master's course)
• Optimization-based Control of Legged Robots (12-hour PhD course)
• Task-Space Inverse Dynamics (3-hour PhD course)

Current PhD advisees include Mohammad Hasan Yeganegi (generalization bounds for imitation learning), Pietro Noah Crestaz (numerically-efficient RL), Veronica Campana (ergodic control for defect detection), Elisa Alboni (data-efficient model-based RL), and Gianni Lunardi (MPC for legged locomotion).