• A. M. Rauf, B. G. Kilberg, C. B. Schindler, S. Park, and K. S.J. Pister. “Towards Aerodynamic Control of Miniature Rockets with MEMS Control Surfaces,” MEMS, Jan. 18-22, 2020. Accepted for publication. PDF.

  • A. Rauf, “New light-based switches could dramatically improve internet speeds,” Aug. 18, 2019. Ranked for a month as the top electrical engineering story on the platform. Link.

Poster Presentations

  • A. Rauf, B. Kilberg, S. Park, and K. Pister. “MEMS Aerodynamic Control Surfaces for Miniature Rockets,” poster presented at Undergraduate Research Symposium, Berkeley, CA, USA. Oct. 8, 2019. PDF.

  • B. Kilberg, A. Rauf, and K. Pister, “Miniature Autonomous Rockets,” poster presented at Berkeley Sensor & Actuator Center Research Review, Berkeley, CA, USA. Sept. 18-19, 2019. PDF.

UC Berkeley

Swarm Lab - Microrockets

This is my current research lab, which I've been working in since January 2019 under Professor Kristofer Pister.

Pico-air vehicles (PAVs), sub-5cm aerial vehicles, are becoming more feasible due to advances in wireless mesh networks, millimeter-scale propulsion, battery technology, and MEMS control surfaces. Our goal is to develop an aerodynamic MEMS control surface that could be used in PAV applications. This device uses electrostatic inchworm motors to rotate a thin silicon fin 15 degrees, which will be used as an airfoil to generate lift as the rocket flies.

In order to determine the aerodynamic performance of the device, we integrated the control surface into a force-sensing platform and operated the device in 23 m/s of airflow. The actuated control surface generated between 0 and 0.25 mN of aerodynamic lift. In order to power this actuator on an untethered PAV, we designed a compact, LiPo battery- powered 90 V power supply PCB that fits in a 3.8 cm x 1.5 cm footprint. We also designed a 20-cm long rocket with onboard power, inertial guidance, and feedback control that we will use as a test platform for the MEMS control surfaces.

I worked on multiple parts of the rocket to integrate the whole system together. I began with a redesign of the MEMS control surface to improve the output force from 2.8 mN to 5.2 mN and to improve the control surface's overall dimensions from 8.8 x 6.7 mm to 5.2 x 8.3 mm, which allows the surface to fit on much thinner rockets than before.

I also worked on shrinking down the overall length of the rocket from 25.5 cm to 18.2 cm through improved friction-fit connectors between components and through leading a PCB redesign of the 3.7V to 80V high voltage converter that drives our inchworm motors off battery power. I've also laid out a path to a 92 cm rocket that I plan to complete by the end of the year, and with custom miniature rocket engines we hope to scale that down to a true 5-cm rocket.

The new PCB's are shipping now, I'll upload a photo once they've arrived

The fully assembled rocket on a custom 3D printed mount

I'm also getting trained in the Marvell Nanofabrication Laboratory on how to assemble these MEMS devices in the cleanroom! To the left is a picture of me, fully gowned-up on my first day in the cleanroom.

You can find the full project description on the Berkeley Sensors and Actuator Center (BSAC) website here.

Devices Physics Lab - Spintronics

From January 2018 to December 2018, I worked with Professor Sayeef Salahuddin in his device physics lab. I worked on two different projects with two different graduate students during my time here.

My favorite project during my time in the group was designing a test structure to measure the DC voltage generated by Al/Py/Pt spin pumping. You can see the three masks I designed below, which were meant to provide a baseline for our group to understand how much voltage we could expect from spin pumping.

The design for the main waveguide

The design for the Permalloy (Py) stripe that provides the isolation needed for spin pumping

The design for the platinum stack

However, after the graduate student I was doing spintronics with graduated, I picked up a project to design a quadrupole magnet testing station to generate magnetic fields up to 150 mT with controllable XY orientation.

I noticed the trouble a number of graduate students were having with our current testing setups, which at the time could only generate highly precise magnetic fields in one direction. While we did have a similar machine with controllable XY magnetic field orientation, it was less precise, required more setup, and was generally meant for larger devices (not to mention how the machine itself measured a meter across).

I thus developed the quadrupole magnet testing station, which you can see to the left. It's composed of two metal bars made of vim var core iron, which have a great rated permeability of 10,000. We passed a current through loops of wire wrapped around the four vertical beams supporting the top plate, which induced a magnetic field through the small opening seen at the top. Designing a feedback controller for the current with LabVIEW gave us very fine control of the magnetic field strength in both the X and Y directions, and the top enclosure was designed to minimize fringing effects for a nearly uniform magnetic field regardless of the net XY orientation.

We fabricated it in the Cory Hall Machine Shop in December 2018, and it turned out quite functional.

As an offshoot of my previous spintronics research, I also happened to pick up a small side project developing a container for MOKE (Magneto-Optical Kerr Effect) testing, which was added to the main testing bench in the lab. You can see an image of the different components needed for the testing setup to the right.

Berkeley Artificial Intelligence Research Lab (BAIR) - Machine Learning

This was my first ever research position, which I worked on from January 2018 to May 2018. I worked with Professor Jitendra Malik and his graduate student Ke Li, and the position taught me a lot about research work culture and presenting my work in a more professional setting. I took up this position together with my first semester of research in Professor Salahuddin's lab since at the time my work with Professor Salahuddin was characterized more like a low-time commitment side project rather than a full research project. However, after a semester of doing both projects, I decided I loved the electrical engineering and device physics side more than my machine learning research, and I focused on taking my spintronics research as far as I could.

I primarily worked on expanding their current algorithms for progressive training of generative adversarial networks (GANs) to reduce their training time, especially with high-dimensional data sets. I also added additional checks to remove potential sources of instability in the neural net's weights as the generative and adversarial networks traded information back and forth between each other.