At RSS 2026 in Sydney, the Test of Time Award was given to a 2014 paper on the RBO Hand, a soft pneumatic robotic hand with no sensors and minimal cost. The award signals a major shift in embodied AI: morphology as computation. The RBO Hand uses physical compliance to automatically adapt to objects, requiring identical control commands for eggs, tofu, or irregular parts. This eliminates the need for complex 3D reconstruction or force-closure algorithms.
For startups, the implication is clear: hardware can solve problems that software alone cannot. The RBO Hand costs pennies to build and repair with $5 glue, contrasting with rigid hands costing tens of thousands of dollars. This cost advantage is critical for scaling in real-world applications where durability and affordability matter more than precision.
The paper's core insight is "Environmental Constraint Exploitation" (ECP): robots should actively use physical contact with surfaces to simplify control, rather than avoiding obstacles. This approach reduces reliance on expensive sensors and compute. The RBO Hand's success in in-hand manipulation without any feedback loops demonstrates robustness that pure software approaches struggle to match.
However, the award also highlights limitations. The RBO Hand operates at low speeds and is not designed for dynamic tasks like running or jumping. For startups targeting high-speed manipulation or locomotion, the morphology-as-computation approach may not yet be ready. The award is a validation for certain niches—dexterous grasping in structured environments—but not a universal solution.
Founders should view this as a strategic signal: invest in hardware innovation that offloads computation to physical design. This is especially relevant for companies building home robots, warehouse grippers, or medical devices where cost and reliability are paramount. The trend also suggests that VC dollars flowing into massive data collection and large models may need rebalancing toward mechanical ingenuity.
The RBO Hand's journey from 2014 to 2026 shows that foundational research can take a decade to bear fruit. For startups, the takeaway is to focus on robust, simple designs that exploit physics, rather than chasing the latest AI trend. The award is a reminder that in embodied AI, the body is as important as the brain.
Source: 雷峰网.