In a groundbreaking advancement, researchers at MIT’s Computer Science and Artificial Intelligence Laboratory (CSAIL) have leveraged artificial intelligence to significantly enhance the jumping and landing skills of robots. By employing generative AI models, the team has developed a robot capable of leaping 41 percent higher than its predecessors, marking a substantial leap forward in robotic design and functionality.
The AI-generated designs feature curved linkages resembling drumsticks, a stark contrast to the traditional straight and rectangular components. This innovative approach allows the robot to leap an average of two feet, a feat previously unattainable with conventional designs. The robots, crafted from polylactic acid, initially appear flat but transform into a diamond shape when activated, showcasing the transformative potential of AI in robotics.
Revolutionizing Robotic Design with AI
This development follows MIT CSAIL’s novel application of generative AI, which allows users to create 3D models of robots and specify parts for modification. The AI then optimizes these areas and tests them in simulation, culminating in a real-world robot fabricated via 3D printing. This process eliminates the need for further adjustments, streamlining the design and production phases.
Generative AI models, similar to OpenAI’s DALL-E, are increasingly instrumental in conceptualizing new designs. These models can generate images, videos, or blueprints, offering fresh perspectives that might not have been previously considered. The integration of AI in robotic design is not just a theoretical exercise but a practical tool for innovation.
AI’s Role in Enhancing Robotic Performance
To refine their jumping robot, researchers sampled 500 potential designs using an initial embedding vector, a numerical representation guiding the AI model’s output. After selecting the top 12 designs based on simulation performance, they optimized the vector through iterative processes, ultimately producing a design that improved the robot’s jumping capabilities.
According to Byungchul Kim, co-lead author and CSAIL postdoc, diffusion models offer unconventional solutions that enhance robotic designs.
“We wanted 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. “However, such a thin structure can easily break if we just use 3D printed material. Our diffusion model came up with a better idea by suggesting a unique shape that allowed the robot to store more energy before it jumped, without making the links too thin.”
Improving Landing Stability
In addition to jumping, the team focused on optimizing the robot’s landing stability. By drafting an optimized foot design, they achieved an 84 percent improvement in the robot’s ability to land safely. This success underscores the potential of AI to enhance not only jumping but also the overall stability and functionality of robots.
Future Implications and Applications
The implications of this research extend beyond jumping robots. Companies involved in manufacturing or household robotics could adopt similar AI-driven approaches to enhance their prototypes, reducing the time and resources typically required for iterative design changes.
Co-lead author and MIT CSAIL PhD student Tsun-Hsuan “Johnson” Wang envisions broader applications for generative AI in robotics.
“We want to branch out to more flexible goals,” says Wang. “Imagine using natural language to guide a diffusion model to draft a robot that can pick up a mug, or operate an electric drill.”
The team is also exploring the potential of adding more motors to control the direction and stability of the robot’s jumps, further enhancing its capabilities. The researchers’ work, supported by the National Science Foundation and other institutions, was presented at the 2025 International Conference on Robotics and Automation, highlighting the ongoing evolution of AI in robotic design.
As AI continues to push the boundaries of what is possible in robotics, the future holds exciting possibilities for more advanced, efficient, and versatile machines. The balance between innovation and practical application remains a focal point for researchers, driving the next wave of technological breakthroughs.
