A Helping Hand on the Wrist

Human-robot interactions have become a key part of the modern industrial landscape, transforming manufacturing processes and expanding into other areas. In the manufacturing sector, these interactions cover a range of collaborative activities, from simple task sharing to complex symbiotic workflows. The integration of robots into manufacturing lines has improved efficiency, quality, and safety, while also highlighting the unique roles that humans play in supervision, decision-making, and adaptability.

But along with these new efficiencies come some concerns. One major worry is the risk of physical harm to human workers due to the close proximity of powerful and sometimes unpredictable robotic systems. Accidents, such as collisions or entanglements, can occur, particularly if safety measures and protocols are not rigorously implemented and followed. Additionally, the complexity of some robotic operations may make it difficult for human workers to anticipate the actions of these machines, leading to potential misunderstandings or miscommunications that could jeopardize both human safety and task performance.

One solution that can help to overcome these safety concerns involves the use of a motion capture system to track the movements of the individual working with the robot. This information can be compared against data from the robot’s motion planning system to provide alerts when a collision is likely, or even shut down the robot to prevent an accident before it happens.

Unfortunately, these systems can be very expensive, sometimes costing hundreds of thousands of dollars. This is because of the advanced technology that is required to deal with situations in which parts of the person become occluded, rendering them totally or partially invisible to the sensing system.

An inexpensive solution to the problem of occlusion has recently been described by a team at the Skolkovo Institute of Science and Technology in Russia that could increase safety in human-robot interactions by making motion capture systems more common. They have developed a lightweight and low-cost forearm-worn robot that can move itself as needed to ensure that it remains in view of a nearby camera-based tracking system. If the wearer’s arm comes too close to the robot it is interacting with, haptic feedback is provided as a warning.

The wearable robot, named ArUcoGlide, straps to the forearm and has a movable platform with an ArUco marker attached to it. The vision system recognizes the ArUco marker to track the location of the wearer’s arm. Powered by an ESP32 microcontroller and two SG90 servo motors, the position of the platform is adjusted by a base computer (via a Bluetooth connection) when occlusions occur that would otherwise obscure the marker from the view of the camera.

Using a computer vision system, the coordinates of the individual’s arm in three-dimensional space are calculated by the base computer. These coordinates then feed into a collision avoidance system that warns the user, via haptic feedback from a pair of vibration motors, when they get too close to the robot. This unique approach eliminates the need for multiple markers or cameras, and reduces the complexity of the processing algorithm.

A series of experiments were conducted to assess the performance of ArUcoGlide in realistic scenarios. Throughout the course of these trials, it was observed that, on average, the participants using ArUcoGlide kept their hands five centimeters further away from the robot than those that did not. And in the course of running through a scenario in which a robot assisted a human with the common laboratory task of pipetting, the speed of the process was increased by 16%.

In the future, the researchers hope to build a GlideSuit, which would have markers positioned all over the body. Such an upgrade would allow the system to be used in a much wider range of use cases.