Transitioning to humanoids: opportunities and challenges
TL;DR
This paper compares a conventional workcell with fixed robotic arms equipped with parallel grippers, and external vision, against a humanoid robot equipped with three-finger hands and onboard RGB‑D sensing, and assesses the differences in both use cases.
Abstract
Robotic manipulation has traditionally relied on fixed-base robotic arms with parallel grippers and external sensors for structured and repetitive tasks, which could be mounted on mobile platforms. Recent advances in limb-based locomotion and dexterous end effectors are enabling a shift toward mobile humanoid platforms. This paper examines the opportunities and challenges of that transition based on the state-of-the art in 2026 on the use case of sorting mixed textiles from an unordered heap into designated bins. We compare a conventional workcell with fixed robotic arms equipped with parallel grippers, and external vision, against a humanoid robot equipped with three-finger hands and onboard RGB‑D sensing. We assess the differences in both use cases regarding deployment flexibility, workspace reachability, safety certification, reliability and in-hand manipulation. As a preliminary step, we define a perception–manipulation pipeline for deformable objects and identify key bottlenecks that currently limit performance: category classification, grasp stability on textiles and whole-body motion planning. We conclude with an outlook on the maturity of current humanoid systems and outline the technical and regulatory steps needed for their integration into textile sorting and related industrial domains.