Martin Hjelm

Papers

Total Citations

197

H-Index

About

Martin Hjelm is a robotics researcher whose work lies at the intersection of machine learning and robotic manipulation, with a particular focus on grasp planning and knowledge transfer. His research addresses one of robotics' most challenging problems: enabling robots to grasp and manipulate objects intelligently across diverse tasks and contexts. Hjelm's most notable contributions include developing frameworks for cross-task, cross-object grasp transfer, allowing robots to apply learned grasping knowledge to new objects and tasks — a critical capability for real-world deployment. His 2014 work tackled both the challenge of determining whether an object affords a given task and planning appropriate grasps accordingly. Complementing this, his research on learning human priors for task-constrained grasping incorporates human knowledge to guide robotic manipulation more naturally and effectively. His methodological contributions extend to representation learning, including a Bayesian non-parametric approach for sparse summarization of high-dimensional robotic grasping data, offering a principled way to compress and encode complex manipulation datasets. His work presented at the International Conference on Robotics and Automation has garnered 167 citations, reflecting meaningful influence within the robotics community. Hjelm's research ultimately advances the goal of robots that can generalize manipulation skills flexibly and intelligently.