A Robot Butler in a Day

We’re all still waiting for the robotic butlers promised for decades that never seem to arrive. Sure, we have robots that do a decent job of vacuuming the floors, but what about everything else? Why do we still have to cook dinner and fold laundry for ourselves in this day and age?

Roboticist Nick Maselli has a client who can’t wait any longer to remove laundry folding from their list of chores. So, Maselli was tasked with building a laundry-folding robot in just 24 hours. He accepted the challenge, and immediately got busy trying to make the future arrive at lightning speed.

The resulting machine looks like something out of a low-budget science fiction movie: a mobile cylindrical robot topped with a dome-shaped head and equipped with two articulated arms. Most of the robot’s body parts, including the arm segments and exterior panels, came from a 3D printer. This allowed for rapid prototyping, which was a key advantage — parts could be redesigned and reprinted within hours.

Running through the center of the robot is a vertical Z-axis actuator, allowing the arms to raise and lower. This gives the machine enough reach to pick up laundry from the floor or work on a tabletop. Each arm contains multiple motors that must be precisely aligned so the grippers can coordinate movements when manipulating fabric, a task far more complex than picking up rigid objects.

The robot’s electronics are concentrated around a Raspberry Pi 5 single-board computer that acts as the brain. Despite its small size, the computer handles four camera feeds, motor drivers, a display, speakers, and a microphone. Power comes from a 12-volt, 10-amp-hour lithium iron phosphate battery, chosen for safety and long service life. A custom-designed power distribution PCB and voltage converters regulate and deliver the exact voltages required by the servo motors and control electronics.

Rather than relying on a prewritten motion sequence, the robot learns. First, a human operator remotely controls the arms through a teleoperation system to demonstrate how towels should be folded. These recorded motions become training data for an AI model, which was trained overnight using powerful GPUs. Once uploaded back onto the robot, its vision system — built from four cameras — allows it to locate a towel, determine orientation, grasp it, and perform a sequence of folds autonomously.

Maselli even designed the system so users can upload new AI models, potentially teaching the robot entirely different household tasks. The long-imagined home helper may not cook dinner yet, but at least one tedious chore has finally met its mechanical replacement.