Diffusion models like OpenAI’s DALL-E have gotten increasingly useful in helping brainstorm recent designs. Humans can prompt these systems to generate a picture, create a video, or refine a blueprint, and are available back with ideas they hadn’t considered before.
But did you understand that generative artificial intelligence (GenAI) models are also making headway in creating working robots? Recent diffusion-based approaches have generated structures and the systems that control them from scratch. With or with no user’s input, these models could make recent designs after which evaluate them in simulation before they’re fabricated.
A brand new approach from MIT’s Computer Science and Artificial Intelligence Laboratory (CSAIL) applies this generative know-how toward improving humans’ robotic designs. Users can draft a 3D model of a robot and specify which parts they’d prefer to see a diffusion model modify, providing its dimensions beforehand. GenAI then brainstorms the optimal shape for these areas and tests its ideas in simulation. When the system finds the proper design, you’ll be able to save after which fabricate a working, real-world robot with a 3D printer, without requiring additional tweaks.
The researchers used this approach to create a robot that leaps up a mean of roughly 2 feet, or 41 percent higher than the same machine they created on their very own. The machines are nearly equivalent in appearance: They’re each product of a style of plastic called polylactic acid, and while they initially appear flat, they spring up right into a diamond shape when a motor pulls on the cord attached to them. So what exactly did AI do otherwise?
A more in-depth look reveals that the AI-generated linkages are curved, and resemble thick drumsticks (the musical instrument drummers use), whereas the usual robot’s connecting parts are straight and rectangular.
Higher and higher blobs
The researchers began to refine their jumping robot by sampling 500 potential designs using an initial embedding vector — a numerical representation that captures high-level features to guide the designs generated by the AI model. From these, they chose the highest 12 options based on performance in simulation and used them to optimize the embedding vector.
This process was repeated five times, progressively guiding the AI model to generate higher designs. The resulting design resembled a blob, so the researchers prompted their system to scale the draft to suit their 3D model. They then fabricated the form, finding that it indeed improved the robot’s jumping abilities.
The advantage of using diffusion models for this task, in response to co-lead creator and CSAIL postdoc Byungchul Kim, is that they’ll find unconventional solutions to refine robots.
“We desired to make our machine jump higher, so we figured we could just make the links connecting its parts as thin as possible to make them light,” says Kim. “Nonetheless, such a skinny structure can easily break if we just use 3D printed material. Our diffusion model got here up with a greater idea by suggesting a novel shape that allowed the robot to store more energy before it jumped, without making the links too thin. This creativity helped us learn in regards to the machine’s underlying physics.”
The team then tasked their system with drafting an optimized foot to make sure it landed safely. They repeated the optimization process, eventually selecting the best-performing design to connect to the underside of their machine. Kim and his colleagues found that their AI-designed machine fell far less often than its baseline, to the tune of an 84 percent improvement.
The diffusion model’s ability to upgrade a robot’s jumping and landing skills suggests it might be useful in enhancing how other machines are designed. For instance, an organization working on manufacturing or household robots could use the same approach to enhance their prototypes, saving engineers time normally reserved for iterating on those changes.
The balance behind the bounce
To create a robot that would jump high and land stably, the researchers recognized that they needed to strike a balance between each goals. They represented each jumping height and landing success rate as numerical data, after which trained their system to seek out a sweet spot between each embedding vectors that would help construct an optimal 3D structure.
The researchers note that while this AI-assisted robot outperformed its human-designed counterpart, it could soon reach even greater recent heights. This iteration involved using materials that were compatible with a 3D printer, but future versions would jump even higher with lighter materials.
Co-lead creator and MIT CSAIL PhD student Tsun-Hsuan “Johnson” Wang says the project is a jumping-off point for brand new robotics designs that generative AI could help with.
“We would like to branch out to more flexible goals,” says Wang. “Imagine using natural language to guide a diffusion model to draft a robot that may pick up a mug, or operate an electrical drill.”
Kim says that a diffusion model could also help to generate articulation and ideate on how parts connect, potentially improving how high the robot would jump. The team can also be exploring the opportunity of adding more motors to regulate which direction the machine jumps and maybe improve its landing stability.
The researchers’ work was supported, partially, by the National Science Foundation’s Emerging Frontiers in Research and Innovation program, the Singapore-MIT Alliance for Research and Technology’s Mens, Manus and Machina program, and the Gwangju Institute of Science and Technology (GIST)-CSAIL Collaboration. They presented their work on the 2025 International Conference on Robotics and Automation.
