Quick Overview
In this short showcase video, Matt Wolfe introduces PhoneBot, an experimental robotics project from UCLA's robotics lab. Wolfe breaks down the engineering behind turning low-cost discarded smartphones into autonomous bipedal humanoids. The brief review highlights the system's architecture, balance recovery, and open-source accessibility.
Key Points
- 1.PhoneBot is a bipedal walking robot developed at Dennis Hong's robotics lab at UCLA that uses an old smartphone as its primary processor and sensor system.
- 2.Rather than serving as a remote control, the smartphone's camera acts as the robot's eyes, its internal motion sensors control balance, and the locomotion AI runs locally on the phone.
- 3.The neural network governing locomotion was trained in simulation, where early iterations learned to walk with a limp, before being deployed onto physical hardware.
- 4.The bipedal frame can autonomously stand back up after being laid flat on its face.
- 5.PhoneBot relies on phones costing between $50 and $80 alongside a body costing roughly $400, offering an open-source alternative to research humanoids that typically cost between $3,500 and $14,000.
Summary
Matt Wolfe highlights PhoneBot, a project originating from professor Dennis Hong's robotics research laboratory at the University of California, Los Angeles. PhoneBot converts inexpensive, repurposed smartphones into the central computing unit of a walking bipedal robot. The phone does not function as a handheld controller or remote operator. Instead, the device physically mounts directly onto the robotic chassis to serve as its brain. The phone's built-in camera provides visual input, its internal inertial measurement unit and motion sensors handle real-time balance calculations, and the artificial intelligence system that dictates walking runs on-device.
The robotic locomotion system was first developed and refined inside an artificial simulation. Wolfe notes that during early stages of the simulated training process, the walking policy resulted in a gait that had a pronounced limp. After fine-tuning the model in the virtual simulator, the team transferred the software onto physical phone hardware. Demonstrations show the assembled unit walking upright across living room floors. In addition to basic forward movement, the robot features self-recovery behavior; when knocked over or placed completely flat on its face, the mechanical joints articulate to push the robot back into an upright standing posture autonomously.
A central focus of the PhoneBot project is affordability and accessible research. While commercially available research humanoids commonly cost anywhere from $3,500 to $14,000, PhoneBot operates with a smartphone valued between $50 and $80 alongside physical mechanical parts that cost about $400. To support widespread development, the UCLA team is open-sourcing the design files, providing full access to the required parts list and 3D printing documentation. Although the current iteration is compact and lacks functional arms, it already demonstrates person-following functionality and mobile camera tracking.
Smartphone Integration and Onboard Processing
Developed by Dennis Hong's robotics lab at UCLA, PhoneBot repurposes inexpensive smartphones into autonomous walking robots. Instead of using the handset as an external remote controller, the robot relies on the device itself as its core brain. The phone's camera functions as visual sensors, internal motion hardware provides balance stabilization, and the artificial intelligence models controlling movement execute directly on the smartphone processor.
Locomotion Training and Self-Righting Capabilities
The locomotion AI was developed using simulated training environments, where early algorithmic iterations learned a gait that walked with a limp before stabilizing. Once deployed to physical phone hardware, the bipedal system demonstrates stable navigation across indoor floor surfaces. The mechanical structure also possesses self-recovery locomotion capabilities, allowing it to push itself back into a standing position without assistance after being placed face down on the ground.
Cost Reduction and Open-Source Accessibility
Traditional research-grade humanoid robots range in price from $3,500 up to $14,000, presenting a substantial financial barrier. PhoneBot dramatically lowers this barrier by pairing a $50 to $80 smartphone with a custom body costing approximately $400. The research team is releasing the entire project as open source, distributing the complete parts lists and 3D printing files to allow external recreation.
The Bottom Line
PhoneBot demonstrates that inexpensive legacy mobile hardware can handle the balance, perception, and inference workloads needed for bipedal locomotion at a fraction of standard humanoid research costs. By providing open-source printing schematics and bills of materials, the UCLA project establishes a viable blueprint for low-cost robotic platforms. The video leaves open future design iterations, including the potential addition of functional arms and expanded autonomy tasks.
FAQ
What is PhoneBot and how does it use an old smartphone?
PhoneBot is an open-source bipedal robot created by Dennis Hong's lab at UCLA that uses a cheap smartphone as its onboard brain, using its camera for vision, sensors for balance, and processor for AI locomotion inference.
Where was the artificial intelligence model for PhoneBot initially trained?
The locomotion AI was trained in a computer simulation, where early training versions learned to walk with a limp before the refined policy was loaded onto a physical smartphone.
How does the PhoneBot robot recover after falling flat on its face?
PhoneBot has an autonomous self-righting behavior that allows its motorized leg joints to push itself back up into a full standing position without external help.
What is the estimated cost of the PhoneBot body compared to research humanoids?
PhoneBot uses a $50 to $80 smartphone and a body costing around $400, whereas standard research-grade humanoid robots range from $3,500 up to $14,000.
What project resources are being released as open source for the PhoneBot platform?
The UCLA team is open-sourcing the entire project, including the complete hardware parts list and the 3D print files required to build the robot body.
Worth watching for
Robotics researchers, makers, and hobbyists interested in low-cost open-source hardware, edge artificial intelligence, and creative ways to repurpose legacy mobile devices.
- phonebot
- robotics
- humanoid
- ucla
- open-source
- edge-ai