News

Dr. Karen Dowling’s MSCA-PF was granted for her project titled “Precise X-Y-Z Readout with a micro-Magnetometer Inverted-pyramid Design (PYRAMID), ” in collaboration with Prof. dr. Kofi Mankinwa. Chip-scale magnetometers come in several flavors, the most common being silicon Hall-effect plates that integrate easily with electronics. However, these devices only detect 1D fields, are asymmetric between X-Y and Z directions, and cannot work in extreme temperatures. Karen’s goal is to leverage micromachining  to realize magnetometers with a unique "3D" microstructure that reduces the footprint and improves accuracy. This project will involve the development of the inverted pyramid device through crystallographic etching of <100> CMOS silicon to expose the <111> crystal plane at 54.7°. This enables higher angular accuracy and avoids fabrication misalignment or packaging errors. In parallel, Kofi’s group at TU Delft will develop the CMOS integrated circuit for front-end amplification and switching scheme of the sensor to detect all three components of the field from a singular device. The final year of the project will focus on combining the sensor together with the circuit in an integrated system. This enables new products for 3D navigation in autonomous microsystems such as biomedical implants, power monitoring, and nanosatellites.

Brain-inspired smart devices that can continuously learn from their environment

While smart devices outline strong promises ranging from productivity gains in industry to smart cities and health-monitoring wearables, there is still one major hurdle hindering their successful deployment: long-term robustness. Indeed, once deployed, smart devices are currently unable to autonomously adapt to changes in their environments, new user features, and evolving task requirements. This implies either electronic waste through device replacement, or increased battery drain and maintenance cost for frequent over-the-air device updates. This project will overcome this hurdle by merging the latest neuroscience and machine-learning research in continual learning to endow smart devices with low-power long-term autonomous adaptation.

Six promising young EEMCS researchers receive Veni grant (tudelft.nl)