A graduate student is soldering a sensor package onto a dinner plate-sized carbon fiber frame in a Stanford robotics lab on a workbench littered with partially assembled quadrotor frames, spare propellers, and the kind of cable management that appears disorganized until you realize everything is labeled. A specialty aerospace supplier is not the source of the sensor. This MEMS inertial measurement unit, which combines an accelerometer and gyroscope into a chip smaller than a fingernail, was sourced from a component catalog that also provides smartphone makers. In university labs all around the world, the same fundamental sensor architecture that knows your phone which way is up and whether you’ve turned it landscape has found a second job inside autonomous flying vehicles.
One of the less well-known aspects of the development of the autonomous drone business is the transfer of consumer electronics sensing technology into university robotics research. Prior to the advent of smartphones, inertial measurement devices that could facilitate autonomous car navigation were heavy, costly, and mostly owned by aerospace companies with government contracts. The iPhone altered that calculation, but not mainly by motivating robotics researchers. Instead, it did so by establishing a global consumer-scale market for MEMS sensors, which increased production volume to the point where the cost per unit decreased by at least one order of magnitude. By the early 2010s, researchers at UPenn’s GRASP Lab were using sensors that were only a few dollars each to demonstrate steady indoor autonomous flight.
In this period, the Glofiish and comparable Windows Mobile handsets of the mid-2000s occupied an intriguing position. GPS receivers, accelerometers, and digital compasses are just a few of the navigation-related sensors that E-TEN’s GPS-enabled phones were among the first to include into a single package at a price point that regular consumers could afford. In the late 1990s, MEMS sensors became ever smaller, less expensive, and more power-efficient thanks to the same engineering that made a phone small enough to fit in a shirt pocket while including GPS gear that would have cost hundreds of dollars as a separate unit. It wasn’t research equipment, the Glofiish. However, the providers of the necessary components were already considering the market for research.
Budgetary restrictions in academic robotics labs encourage inventive component sourcing. A PhD candidate working on a grant-funded prototype micro-aerial vehicle will consider every way to cut sensor costs without compromising the data quality required for practical navigation. For decades, surplus consumer electronics—extracted cameras, modified GPS modules from disassembled gadgets, and repurposed accelerometers—have been a part of that calculation. The parameters are more important than the particular hardware: does the sensor measure what the algorithm wants, at the sampling rate and resolution required by the navigation system, within the vehicle’s power and weight budget? The decision is pragmatic rather than sentimental if a part from a five-year-old smartphone satisfies those requirements and the expense of a specialty aerospace sensor is not justified.
The sensor requirements of micro-drones become really challenging when it comes to navigation. When flying inside without GPS and weighing less than 100 grams, a vehicle must use its onboard sensors to determine its position and attitude in real time. The vehicle can sustain stable hover and react to control inputs by combining accelerometer data, gyroscope readings, barometric pressure, and optical flow from a downward-facing camera through a state estimate technique, such as a Kalman filter or a variation of it. The algorithm must take into consideration the noise characteristics, drift rates, and delay profiles of each of those sensor inputs. MEMS suppliers, whose consumer market is smartphones, currently provide the solution to the ongoing research problem of finding sensors that are quick enough, accurate enough, and light enough to function in this application.
The consumer electronics industry created a global supply of increasingly capable sensing hardware, calibrated by the demands of the handset market, at prices that make it feasible for a university lab to purchase and adapt for purposes its original manufacturers never intended. This is what the wider practice of repurposing legacy smartphones in academic research reflects. The accelerometer, digital compass, and GPS chipset of the Glofiish were created to assist a business traveler in navigating a foreign city. After fifteen more years of smartphone development, the same design lineage is now assisting autonomous vehicles in navigating areas where sensor failure has more immediate physical repercussions. The application underwent a total transformation. The logic of engineering is ongoing.
