Stabilizing a High-Power Quadcopter

Overview

Objectives/Goals I was a member of a Stanford UAV club project that built an ultra-high-powered quadcopter with a thrust-to-weight ratio of 8:1 (0 to 60 mph in 0.34 seconds). We hoped to break a world record for fastest quadcopter. Because the quad was constructed from off-the-shelf parts that were known to work well, we expected it to just fly. But instead it just crashed immediately. I was responsible for getting the quad to fly. I had to modify the stock firmware of the flight controller and the electronic speed controls (ESCs) in order to stabilize such a high-powered quadcopter. Methods/Materials I came up with the following procedure for determining and fixing the cause of the crashes. First, I lowered the power of the quad (by modifying the ESC firmware) until the quad flew stably. Then, in successive flights, I gradually increased the power until I observed a problem with the quad's flight behavior. Once I identified a problem, I modified the flight controller firmware to eliminate it. Then I continued to increase the power so I could observe any more problems. Results I identified and repaired the following problems with the flight controller's firmware. First, I had to fix a bug that caused an uncontrollable throttle increase during a hover. Second, I had change the firmware to allow PID constants less than 0.001. The third problem required a particularly involved solution. The flight control firmware used a single set of PID gains for the entire throttle range. Unfortunately, my flight testing showed that a single set was insufficient for a high-power quadcopter. So I modified the firmware to implement a 10-segment piecewise-linear gain curve. I devised a novel procedure for determining the required gains at 10 points in the throttle range. Conclusions/Discussion I concluded that conventional flight control software can't be used to fly a high-power quad stably, without modifications. The methodology I devised: starting with conventional power levels and gradually increasing the power, allowed me to identify and fix problems in the flight control firmware. I showed that PID gains need to vary with the throttle, and a piecewise-linear curve can implement this variation. My experiments for determining the shape of this curve yielded a surprising result: the piecewise-linear curve is U-shaped rather than monotonically decreasing as is widely believed.

Competition history

CSEF 2016 Applied Mechanics & Structures · Entry S0313

Resources

CSEF project summary

Related projects

ISEF · 2016