Thomas Chabal

Centre National de la Recherche Scientifique, Institut national de recherche en sciences et technologies du numérique

Papers

Total Citations

H-Index

About

Thomas Chabal’s research sits at the intersection of embodied AI, robotic navigation, and physical reasoning. His most cited work, “Object Goal Navigation with Recursive Implicit Maps” (2023, 11 citations), tackles a core challenge in autonomous robotics: enabling agents to locate specific object categories in unfamiliar environments. Rather than relying on traditional explicit mapping—which demands heavy engineering and lacks semantic depth—Chabal introduces a recursive implicit mapping approach that fuses geometric and semantic information, allowing for more efficient and intelligent object-oriented exploration. This work bridges the gap between classical map-based methods and modern end-to-end learning, offering a scalable solution for real-world navigation tasks. In “Assembly Planning from Observations under Physical Constraints” (2022, 2 citations), he extends his focus to manipulation, developing algorithms that can infer and replicate unknown physical assemblies from a single photograph, using object detection and pose estimation under real-world stability constraints. Chabal’s contributions are particularly valuable for students and researchers interested in how robots can understand and interact with unstructured environments without exhaustive prior knowledge. His work demonstrates a clear trajectory toward more autonomous, semantically aware robotic systems.