Maria Koskinopoulou, a robotics and computer vision researcher at Heriot-Watt University, discusses sensing and manipulation for medical robotics and recycling with Robot Talk host Claire Asher.
Koskinopoulou’s lab combines robotics, vision and machine learning to help robots handle objects in settings such as healthcare and recycling. Its work includes designing grippers for delicate objects and teaching robots to take apart electronics such as hard drives and battery packs, where tiny hidden parts and variable materials make the task difficult. She says combining vision and force sensing can help robots apply learned skills to unfamiliar devices.
Koskinopoulou describes sensing and manipulation as a continuous feedback process: robots need to assess an object and adjust as they interact with it. Her research includes learning from demonstration, where robots acquire skills by observing examples and practicing with their environment. Touch and force feedback can help them respond when conditions change, such as when an object slips.
Her medical robotics work includes needle-based procedures such as blood sampling, catheter placement and thoracentesis. A system combines ultrasound imaging, robotic movement and autonomous control to locate a target and assist with needle insertion, while responding to movement in tissue and organs. In teleoperated systems, force feedback through joysticks can help clinicians sense obstacles and differences in tissue.
Koskinopoulou says most robotic surgery systems are teleoperated, with the surgeon directing the procedure; autonomy is better understood as a range of levels than an all-or-nothing choice. She sees medical robots as tools to support clinicians, while researchers test smaller autonomous tasks in controlled settings. One example combines near-infrared imaging to locate veins with a bioimpedance sensor that detects blood during needle insertion.
Recycling facilities present crowded, moving streams of dirty, crushed and overlapping materials. Koskinopoulou describes robotic arms using computer vision and sometimes multispectral cameras to identify items and sort them, often with suction grippers to isolate objects. Her team deployed a sorting system with an industry partner in Greece, and she says models need ongoing data collection and retraining to handle different waste streams and regional differences.
Briefs summarise talks and interviews; timestamps open the video at that moment. Claims in a talk are the speaker's own.
Rika Antonova, associate professor at the University of Cambridge, discusses co-evolving robot hardware and reinforcement learning algorithms with Robot Talk host Claire Asher.
Aaron Edsinger, Hello Robot’s CEO and co-founder, talks with Claire Asher on Robot Talk about designing practical assistive robots, safety, and bringing Stretch into homes.
Allison Okamura, a Stanford University engineering professor, discusses human-centered robotics, haptic feedback, wearable devices, and soft robots for medicine, learning, and hazardous environments with Robot Talk host Claire Asher.
Josie Hughes, assistant professor at EPFL and head of the CREATE Lab, discusses AI-assisted robot design, soft robotics, and reconfigurable machines with Robot Talk host Claire Asher.
Rich Walker, director of Shadow Robot Company, discusses dexterous robot hands, their design trade-offs, and the challenges of bringing robotics into practical use with Robot Talk host Claire Asher.
Amir Patel, associate professor at University College London, discusses how robotics can reveal how cheetahs move and how those findings inform agile robots, wildlife monitoring, and animal health research with Robot Talk host Claire Asher.
