Mathilde Kappel

Centre National de la Recherche Scientifique

Papers

Total Citations

H-Index

About

Mathilde Kappel is a roboticist advancing the frontier of dexterous manipulation through quality-diversity algorithms. Her research centers on generating diverse, high-quality robotic grasps—a critical bottleneck for generalizable robot learning. In her 2024 work, "Speeding up 6-DoF Grasp Sampling with Quality-Diversity," Kappel demonstrated how evolutionary optimization can efficiently produce a rich repertoire of grasp poses, accelerating the sampling process for six-degree-of-freedom manipulation. This approach directly addresses the data scarcity that limits generalization in robotic grasping. Building on this, she introduced QDGSET (2025), a large-scale synthetic grasping dataset generated using quality-diversity methods. Unlike prior datasets created with simple sampling techniques, QDGSET offers unprecedented diversity and quality, providing a robust foundation for training grasp prediction models. Though early in her career, Kappel’s work has already garnered attention (over 5 citations across her key papers), and her datasets are poised to become standard benchmarks. By fusing evolutionary computation with robotic learning, she is helping to close the interaction-data gap—paving the way for robots that can grasp any object, anywhere.