Robots don’t have to be expensive. In fact, if there’s a smartphone handy, a wheeled ‘bot can be built for around $50, according to Matthias Muller and Vladlen Koltun. The two Intel researchers are the coauthors of a recent paper titled “OpenBot: Turning Smartphones into Robots,” which proposes leveraging phones to equip robots with sensors, computation, communication, and access to an open software ecosystem.
Using off-the-shelf smartphones as robot brains confers a number of advantages beyond cost savings. Rapid upgrade cycles mean secondhand smartphones are readily available — an estimated 40% of the world’s population owns smartphones — and they’re constantly improving with regard to camera quality and processor speed. Moreover, even commodity models sport inertial measurement units, GPS, Wi-Fi, Bluetooth, cellular modems, and dedicated chips for AI inference, some of which outperform desktop processors.
In the researchers’ OpenBot design, a smartphone plugs into an electromechanical body for sensing, data fusion, and computation. The robot’s chassis is designed to be 3D-printable and accommodate up to four motors, with cutouts for a controller, microcontroller, LEDs, a smartphone mount, and a USB cable. A battery pack connects to dedicated terminals to deliver power to the motors as needed, and an Arduino Nano board pairs with the smartphone via USB, providing a serial communication link and power. The two front wheels are equipped with sensors that send odometry signals, and pins connected to the motor controller allow for real-time speed and direction adjustments.
OpenBot’s software stack is a bit simpler, consisting of only two components that communicate via the serial link. An Android app runs on the smartphone, providing an interface for an operator to collect data sets while running higher-level perception and control workloads. Meanwhile, a program running on the Arduino takes care of low-level actuation, as well as measurements like odometry and battery voltage.
Courtesy of the Android app, OpenBot can be controlled with off-the-shelf Bluetooth-compatible PS4, Xbox, and Switch game controllers. The buttons on the controllers can be assigned to things like data collection and switching among path-finding models, including out-of-the-box models for autonomous navigation that detect and follow people within the robot’s line of sight.
Muller and Koltun conducted tests to compare the robot’s performance with various smartphone models. While the cheapest low-end phone they tested (a Nokia 2.2) performed the worst, it still managed to detect and follow a person about half the time. Moreover, all of the recent mid-range phones tested (including the Xiaomi Note 8, Huawei P30 Lite, and Xiaomi Poco F1) tracked people consistently at 10 frames per second or higher thanks the phones’ dedicated AI accelerators.
“This work aims to address two key challenges in robotics: accessibility and scalability. Smartphones are ubiquitous and are becoming more powerful by the year. We have developed a combination of hardware and software that turns smartphones into robots. The resulting robots are inexpensive but capable,” the researchers wrote. “Our experiments have shown that a $50 robot body powered by a smartphone is capable of person following and real-time autonomous navigation. We hope that the presented work will open new opportunities for education and large-scale learning via thousands of low-cost robots deployed around the world.”
The audio problem: Learn how new cloud-based API solutions are solving imperfect, frustrating audio in video conferences. Access here