Evaluating humanoids for surface finishing applications

Manufacturing processes such as sanding, polishing, blasting, grinding, coating, and painting are used in a wide variety of industries. These manufacturing tasks are ergonomically challenging and pose significant health risks for humans (please see image above). There is considerable interest in using robots to automate these operations. When parts are large, the robot arm needs to be repositioned around to ensure that the entire part surface can be covered. People often ask if humanoids would be a good option to perform manufacturing operations on large parts. Let us evaluate different humanoid features and examine their utility in manufacturing process operations. Legs: Legs are not efficient for locomotion on flat surfaces found on factory floors. Moreover, we often need to tether the robot to provide power for processing tools (e.g., a grinding disk) used in manufacturing applications. If a legged platform needs to be tethered, then it does not offer much flexibility. So, legs are far from optimal means of locomotion on the factory floor. Fixed rails mounted on the floor or mobile base are a better alternative for transporting robot arms in processing applications. Legs add complexity, increase safety risk without offering any value. Therefore, we do not need legs on factory floors for manufacturing operations. Submit your session idea for the 2026 RoboBusiness Multi-Fingered Hands: Hands with multiple fingers offer flexibility and dexterity in object manipulation. However, hands are expensive. Many manufacturing processes such as sanding, grinding, blasting, coating, screw-driving focus on tool manipulation. Holding a surface finishing tool with a multi-fingered hand may not provide a strong grasp needed to perform the process at a high speed. So, we are better off directly connecting the tool to the arm via a simpler connector and lower hardware cost by removing complex hands. Heads: The head mounted on a humanoid is small, so cameras need to be placed close to each other in the head. If we want good coverage and accuracy, then we need to spread cameras (and other sensors) in the cell. Often sensors mounted on the arm close to the tool can be much more effective. Moreover, cameras can be placed in the cell above part. So, heads do not offer much value in manufacturing applications. Dual Arms: Dual arms are very useful in processing large parts. If we want to increase coverage and finish the part quickly, then we need to place arms sufficiently far away from each other to minimize collision risks with each and maximize the workspace. This is a different arm placement configuration from what is seen on humanoids. The robot configuration that works well on a surface finishing application is depicted in the image below: It uses dual arms. It uses a mobility solution to reposition arms. It uses vision to understand the scene. However, component configuration used here is very different from the component configuration in humanoids. People often argue that humanoids enable economies of scale. The configuration shown above features general purpose robot arms, rails, cameras, and force sensors. Most of the hardware components in this workcell benefit from the economy of scale. Therefore, humanoids are not needed to leverage economies of scale in manufacturing applications. Most manufacturing applications aim to reduce cycle time to improve throughput. So, using general-purpose industrial robots that transcend human constraints is often an attractive option. These robots can operate 4 to 5 times faster than human speeds and can apply significantly higher forces. They are already proven to work reliably in really demanding applications. So building manufacturing workcells by placing industrial robot arms in the right configurations seems to be a winning solution in most manufacturing applications. Therefore, humanoids are not likely to be a winning solution in manufacturing applications where precision, performance, and speed are the main decision criteria. About the author Dr. Satyandra K. Gupta is co-Founder and chief Scientist at GrayMatter Robotics . He also holds Smith International Professorship at the Viterbi School of Engineering at the University of Southern California and serves as the founding Director of the Center for Advanced Manufacturing. His research interests are Physical AI and human-centered automation. He has published more than five hundred technical articles in journals, conference proceedings, and edited books. He also holds twenty eight US patents. He is a fellow of the American Association for the Advancement of Science (AAAS), American Society of Mechanical Engineers (ASME), Institute of Electrical and Electronics Engineers (IEEE), National Academy of Inventors (NAI), Society of Manufacturing Engineers (SME), and Solid Modeling Association (SMA). He currently serves as a member of the T