Adrian Calma

About

Adrian Calma’s research lies at the intersection of robotics, machine learning, and human-robot interaction, with a central focus on making industrial robots more autonomous and efficient. His key contributions revolve around **active learning**—a paradigm where robots intelligently query for the most informative data to train themselves, reducing the need for costly, manual reprogramming. In his most cited work, "Active Sorting – An Efficient Training of a Sorting Robot with Active Learning Techniques" (2018, 6 citations), Calma introduced a probabilistic active learning approach that enables a robot to master object sorting tasks with minimal human intervention. He further advanced the field by tackling the challenge of noisy, real-world data. In "Combining Self-reported Confidences from Uncertain Annotators to Improve Label Quality" (2019, 4 citations), he pioneered a method for leveraging annotators’ own confidence ratings to enhance label accuracy, a critical step for reliable robot training. His work "Automated Active Learning with a Robot" (2018, 2 citations) demonstrates a fully automated learning pipeline, making the process intuitive for human operators. Calma’s research is notable for its practical, industry-oriented approach, directly addressing the demand for flexible, self-adapting automation.

Research Focus

Key Achievements

Top Papers

1. 1
   Active Sorting – An Efficient Training of a Sorting Robot with Active Learning Techniques

   6 citations · 2018
2. 2
   Combining Self-reported Confidences from Uncertain Annotators to Improve Label Quality

   4 citations · 2019
3. 3
   Automated Active Learning with a Robot

   2 citations · 2018

Key Collaborators

Contact & Links