During the spring of 2020, I worked with Rice University and Baylor College of Medicine faculty to develop a low-cost ventilator in response to the COVID-19 pandemic. We aimed to create a device, built using off-the-shelf components, to function as an emergency High Acuity Limited Operability (HALO) ventilator.

At the onset of the pandemic, several emergency medicine faculty at Baylor approached Rice’s Oshman Engineering Design Kitchen (OEDK). They indicated that while practical COVID guidelines were in development, there were far too many patients who required intubation and a ventilator. The faculty at the OEDK reached out to me, as I had been working closely with them previously, and asked for assistance in designing and engineering a device capable of aiding this situation. We called the device the ApolloBVM – a reference that we felt was fitting, considering the project felt like a “moonshot” at the time.

We created an electromechanical device featuring two actuated linear arms to compress a bag valve mask (BVM) – a widespread manual ventilation solution in every U.S. hospital. In essence, the device was a robot with two degrees of freedom. My role was to create a redundant control system and user interface to control the device accurately and efficiently.

From my experience with robotics and motion planning, I designed a motion planning algorithm that – when provided with a respiratory rate, tidal volume, and desired front-end pressure – would control the arms of the device, compress the BVM, and deliver air to a patient. The system ran on two identical Arduino UNOs, a relatively low-cost and abundant microcontroller (MCU) platform. One of these microcontrollers was used to run the user interface, while the other planned motions and tracked the resulting trajectories using a PID control loop.

After calibrating the device using medical-grade test equipment on campus, we ran verification tests to address any reliability concerns. Not only could our ApolloBVM devices provide air to patients with profiles matching that of commercially available ventilators, but they were able to do so continuously for up to 3 weeks. With the aid of our colleagues at Baylor, we applied for and received EUA approval from the U.S. FDA and IRB approval to conduct human trials at Baylor St. Luke’s Hospital.

As the project had intended to be an emergency solution with common components, we created a website to distribute instructions on manufacturing identical devices. We included an open-source repository on GitHub to host the code and run continuous integration tests. As Arduino MCUs are typically points of first contact for embedded software development and not serious embedded frameworks, I opted to write everything in C++ and use the Platform I.O. framework to distribute binaries instead of requiring manual codebase compiling. Furthermore, as only some people who would like to build an Apollo BVM would have access to the necessary manufacturing machines, we collaborated with DesignNest to offer kits for $45 to make our devices.

Regardless of the means of production, we received notes that devices were built in over 115+ countries worldwide, and over 3000 individuals downloaded the plans. While we don’t know how many of those devices saw use in hospitals, it was very impactful to see such widespread adoption. Following this, the project received widespread media attention in local and national news. As of late 2020, we had licensed the device to Stewart and Stevenson to bring ApolloBVM to diverse markets.