Max Robotics — Global Service Robot Directory + US Certification

About

Luigi Berducci is a researcher at the forefront of autonomous systems, specializing in the intersection of deep reinforcement learning (RL), world models, and multi-agent safety. His work is defined by a practical, engineering-driven approach to bridging the gap between theoretical RL and real-world robotics. Berducci’s most impactful contribution, "Latent Imagination Facilitates Zero-Shot Transfer in Autonomous Racing" (35 citations), demonstrates how world models can learn behaviors in a latent space, dramatically improving sample efficiency and enabling a policy trained in simulation to transfer directly to a real autonomous race car without any fine-tuning. This work is complemented by his comparative study of model-based versus model-free RL for autonomous racing (18 citations), which provides crucial empirical insights for deploying these algorithms on physical hardware. More recently, Berducci has tackled the critical challenge of safety in multi-agent systems, introducing an adaptive Control Barrier Function method (6 citations) that relaxes strong assumptions about other agents’ behavior. His work on hierarchical reward shaping from task specifications (4 citations) further showcases his ability to synthesize complex, multi-objective control policies. Berducci’s research is essential reading for anyone interested in deploying robust, sample-efficient, and safe RL agents in the real world.