In January, humanoid robotics company Figure demonstrated its Figure 02 robot at work, sorting packages in a logistics warehouse with its sophisticated Helix visual language system. The video featured humanoid robots easily manipulating packages of various shapes, sizes, and stiffnesses, with a degree of precision and flexibility that was decidedly human.
Three months later, Figure has already published a stunning update. The robot is now able to handle an even wider range of packages, such as deformed poly bags and flat mailers, and do so more quickly and more precisely. The company says that Helix has developed quickly, allowing Figure 02 to sort a package in 4.3 seconds, a full second quicker than previously, and with even greater accuracy.
One of the most impressive developments is that the robot can dynamically change its grip in real time. It is able to flick soft packages to reorient them, pinch flat mailers with precision, and even pat down crumpled plastic to reveal barcodes. This seemingly simple “flattening” action, learned purely from demonstration, wasn’t hard-coded—highlighting the power of Helix’s end-to-end learning system.
This kind of action highlights the rate at which humanoid robots are closing the gap between human abilities in warehouse operations. Robots such as Figure 02 are no longer restricted to repetitive and strictly defined movements. Rather, they are being trained on subtle tactics using data itself, which makes them very flexible to the realities of the logistics work, which is not perfect.
As companies such as Tesla, Agility Robotics, and Figure continue to work to automate additional human jobs, the future is not without its challenges. Nevertheless, the challenge of scaling humanoid robots to be cost-effective, robust, and genuinely competitive with human workers is a major technical and economic challenge, even despite these breakthroughs.
Nevertheless, it is a remarkable vision of the future to see a robot skillfully sorting packages, flattening, flipping, and scanning with a smoothness that is hard to imagine. The warehouse of the future could easily be manned not by humans, but by a relentless, constantly learning crew of humanoid robots.
