One of the biggest challenges in humanoid robotics is not teaching robots how to walk, it is making them walk efficiently.
This is the ambition behind the project led by Andrés González and Alexander Palatnik, winners of the Third Open Call of dAIEDGE.
Their work explores how Bayesian inference on embedded edge systems can enable a new generation of energy-efficient humanoid robots. The project builds upon tendon-based actuation technology originally developed at the Max Planck Institute for Intelligent Systems, extending this research into a practical robotic platform capable of significantly longer operating times.
The results are remarkable. While many current humanoid platforms operate for between one and three hours on a single battery charge, the team's prototype is designed to reach up to 12 hours of autonomy, opening the door to robots that can perform meaningful tasks without frequent interruptions for recharging.
Achieving this requires much more than efficient mechanics. By combining intelligent control algorithms with edge-based inference, the project demonstrates how AI can contribute directly to improving energy efficiency without compromising robotic performance.
Throughout the programe, they collaborated with VERSES, whose mentorship helped guide the project's technical evolution. We would like to sincerely thank VERSES for their commitment as hosting institution and for supporting the team throughout the programme.
The seven-month journey also allowed the team to strengthen and validate their developments through the dAIEDGE-VLab, where they could experiment with their solutions in an environment designed for Edge AI research. Combined with expert mentoring and financial support, this enabled the team to focus entirely on transforming an ambitious concept into tangible progress.
Projects like Next Prototypes e.V. illustrate how innovation in Edge AI extends beyond algorithms, enabling smarter, more sustainable robotic systems that are ready to operate in the real world.
Want to hear the story behind the project? Watch the full interview with Andrés and Alexander clicking here.