Max Robotics — Global Service Robot Directory + US Certification

About

Toon Van de Maele is pioneering the intersection of active inference and robotics, developing systems that perceive and act like living organisms. His research centers on three key areas: hierarchical active inference for robot navigation, active vision for manipulation under constraints, and object-centric scene understanding. Van de Maele's most impactful work, "Robot navigation as hierarchical active inference" (2021, 79 citations), introduces a framework where robots navigate by minimizing surprise, learning spatial hierarchies from raw sensory data. In "Active Vision for Robot Manipulators Using the Free Energy Principle" (2021, 23 citations), he addresses real-world sensing limitations—occlusions, restricted field of view, and resolution—by enabling robots to actively query multiple observations before acting. His 2024 paper on "Object-Centric Scene Representations Using Active Inference" (5 citations) advances scene understanding by allowing agents to infer objects from raw sensory data without supervision. Van de Maele also tackles computational efficiency in robotics, optimizing image-based reinforcement learning for resource-constrained systems (2022, 6 citations). His work bridges cognitive science and robotics, offering principled, biologically-inspired solutions for autonomous agents.