Researchers at the University of California, Los Angeles (UCLA), have introduced PhoneBot, an open-source humanoid robot platform designed to lower the financial barriers associated with bipedal robotics research. According to technical reports and project documentation, the platform utilizes inexpensive or retired Android smartphones to manage core computing, vision, and sensing tasks, bypassing the need for costly onboard computers and dedicated sensor arrays. With core robotic hardware expenses estimated around $400, the project aims to make compact humanoid systems much more accessible for academic laboratories, classrooms, and independent makers.
Hardware Architecture and Smartphone Integration
PhoneBot stands 483 millimeters tall and weighs 1.8 kilograms, operating via a configuration of 13 servo motors that manage its legs and torso movements. Rather than incorporating a dedicated processor and external cameras, the platform mounts a commodity smartphone to its upper torso. The phone's built-in camera functions as the robot's primary set of eyes for human tracking and visual detection, while its integrated inertial measurement unit supplies motion telemetry required for balance and stability.
A secondary microcontroller inside the body communicates wirelessly over Wi-Fi with the smartphone, translating high-level commands into precise joint activations. The lower body features six actuated joints per leg, complemented by an additional torso yaw joint that permits the mounted smartphone to rotate independently. This mechanical arrangement allows the system to visually track individuals or execute perception tasks without requiring the robot to rotate its entire stance.
Reinforcement Learning and Gait Optimization
The robot executes locomotion policies powered by TensorFlow Lite running directly on the smartphone hardware. Training the walking controller involved reinforcement learning simulations designed to reward stable movement while penalizing awkward postures and excessive motor strain. Initial simulation runs yielded uneven gait patterns where the robot favored one leg over the other, presenting a typical hurdle in bipedal simulation-to-real transfer.
To resolve this, researchers introduced mirrored training samples that paired corresponding left and right movements, effectively enhancing gait symmetry without requiring complex additional physics steps. Evaluations demonstrated that this training method successfully reduced joint-angle symmetry errors and minimized left-right torque and mechanical-power imbalances during physical demonstrations, which included walking and recovering from a face-down position.
Democratizing Bipedal Robotics Through Upcycled Technology
The introduction of PhoneBot addresses a fundamental bottleneck in robotics education and experimental research: prohibitive hardware expenses and complex system integration. Historically, small-scale humanoid platforms required specialized sensing suites, custom circuit boards, and high-performance onboard computers that kept deployment costs out of reach for many academic institutions and independent developers. By shifting high-level computation and perception duties to off-the-shelf consumer smartphones—including older models like a 2017 Honor 9 and budget handsets priced between $50 and $80—the UCLA team demonstrates a viable pathway toward hardware democratization.
Despite these cost-saving innovations, the platform retains notable physical limitations that affect its immediate utility. Its low-powered actuators restrict the payload capacity and dynamic agility of the robot, while its lack of arms prevents object manipulation tasks. Furthermore, current locomotion controllers are optimized primarily for flat terrain, leaving uneven-surface mobility and robust outdoor deployment as areas for future investigation. For researchers, educators, and hobbyists, however, PhoneBot provides a fully open-source blueprint to study bipedal locomotion, mobile telepresence, and smartphone-based robotics without investing thousands of dollars into proprietary hardware ecosystems.
Editorial graphic and headline text by Signal Titans.
Source: PhoneBot uses an old $50 Android smartphone brain with its cameras as eyes — tech Frankenstein is an open source humanoid robot platform for makers · PhoneBot: A Low-Cost Open Humanoid Robot Platform Reusing Smartphones
SIGNAL TITANS
Leave a Reply