Chris Cross

Two labs on a Crazyflie nano-quadcopter. First, parameter estimation: we flew a scripted routine, logged 19 states (position, velocity, Euler angles and rates, accelerations and the four rotor speeds), mapped rotor PWM to angular velocity (ω = 0.04 · PWM) and set up a least-squares estimate in MATLAB for the moments of inertia and the thrust and drag coefficients. A sixth parameter, rotor inertia I_r, came out at about −1.1 × 10⁻⁷, which confirmed it is negligible in the flight model.

Second, navigation: we tuned PID gains on x, y and z to keep step-response overshoot under 10% (z: K_p = 2, K_i = 0.5, K_d = 0), then flew a multi-waypoint trajectory. Raising K_p by one destabilized the quad almost immediately. Raising K_d by 0.3 still tracked, but the flight was choppier. Finally we configured proximity-sensor obstacle avoidance in Python, setting the trigger distance and retreat velocity.

Member of a four-person team with Drew, Alex H and Alex K, working across the flight script, the MATLAB least-squares estimation, and the gain tuning and avoidance tests. We also changed the flight script to add a slow ascent and descent so the quad landed gently instead of cutting out mid-air.

• System identification by least-squares regression from logged flight data
• Finite-difference estimation of angular accelerations (T = 0.1 s)
• PID tuning against an overshoot spec, and how sensitive stability is to K_p vs. K_d
• Reading and modifying flight-control Python; safe hardware test practice

SoftDebugging hardware under time pressure, team communication, test planning.

• Crazyflie quadcopter, radio dongle, onboard proximity sensors
• Python flight scripts and PID tuning interface; MATLAB for data and regression
• Lab manual, padded flight area, course TA

Fabrication of a CMOS wafer by cleanroom processing using all six masks: p-well, PMOS source/drain, NMOS source/drain, gate oxide, contact vias, and metal. Each process is recorded on the process traveler along with measurements of the test wafer.

The field and gate oxides are fabricated using wet and dry thermal oxidation. The p-well and both the source/drain regions are fabricated using solid source diffusion of boron and phosphorous, followed by drive in diffusion. The predicted BOE etch times are obtained using calibration of oxide thickness on test wafer before each photolithography process. Aluminum contacts are deposited using e-beam evaporation, while interconnect is patterned by PAN etch forming an ohmic contact.

• Semiconductor fabrication process, from wafer preparation till sintering of metal contacts
• Physics of oxidation and diffusion in determining time and temperature of the processes
• Photolithography, wet etching (BOE, PAN), etch time prediction
• Cleanroom procedures and discipline in documentation

• University clean room facilities: furnaces, mask aligner, wet benches, e-beam evaporator
• Six layer mask set, process traveler, calibration test wafer
• In progress