Mechanical Minds Spring Into Action

These days, programming has pretty well become synonymous with coding in languages such as Python or C++. In the emerging vibe coding scene, programming also involves a whole lot of time spent belittling an AI coding assistant and repeatedly typing “that still doesn’t work, try again.” In days past, more exotic (by today’s standards) programming methods were regularly used; everything from punch cards to patch panels had their time in the sun.

Prior to the digital era, things were even stranger. Mechanical components, like gears, effectively encoded a program into a physical object. For the most part, that is not the way of the world anymore. But there are some advantages to this approach that engineers at Harvard have rediscovered. They have shown that robotic sensing and control systems can be built with little more than carefully tuned tension in rubber bands.

The team has created a robot that can navigate its surroundings and perform simple tasks — like avoiding obstacles or sorting objects — without the need for many conventional electronic components or a software control system. Instead, the intelligence of the robot is built directly into its physical structure.

Each rubber band assigns a different energy cost to the rotation of a lever, and the robot naturally follows the path of least energy. This arrangement causes it to essentially execute a built-in set of mechanical instructions. When a leg is attached to the mechanism, it becomes a walker capable of moving forward, backward, or even adjusting its gait depending on its configuration.

For sensing, the robot “feels” its environment using a pair of antennae. When one of these antennae bumps into an obstacle, the interaction mechanically reprograms the robot’s energy landscape, causing it to turn and move away without any traditional sensors or microcontrollers. The same underlying mechanism was adapted to sort objects based on mass, picking up and dropping them at specific points determined by their weight.

While the prototype is simple, it demonstrates that machines can exhibit autonomous, adaptive behavior through mechanical design alone. That could lead to small, lightweight robots that are cheaper, easier to produce, and more robust than their electronically complex counterparts.

Looking ahead, the researchers envision physically intelligent robots built from flexible materials that can move faster, jump, or operate in environments where electronics would fail. Perhaps sometimes machines don’t need a brain at all. Just a few well-placed rubber bands may be enough to do the trick for some applications.