Email
Password
Remember meForgot password?
    Log in with Twitter

article imageNew app assesses just how ergonomic warehouse jobs are

By Tim Sandle     Aug 24, 2019 in Technology
Technologists have applied machine learning to build a different system to monitor and assess factory and warehouse workers and inform them how ergonomic their jobs are, with the data reviewed and calculated in real time.
Musculoskeletal disorders occur at work when workers need to use awkward postures or where repeated tasks are performed. Such behaviors place a strain on the body, which can become worse over time. As part of best employment practices, employers should set out to minimize those behaviors that will lead to strains or injuries occurring and put in place measures designed to keep workers healthy while performing their duties.
The problem with many assessments is that they take time and the results of ergonomic analysis are rarely available in real-time, which is when the workers need the information. To address this, scientists working at the University of Washington have deployed machine learning to come up with an alternative system to monitor factory and warehouse workers and inform them how risky certain behaviors might be.
This comes from an algorithm that divides up a various activities like lifting a box from a high shelf or carrying a package into individual steps. The program, which is available as an app, can then calculate a risk score based on each action that is required to complete a given task.
As an example, the app assesses each position involved with a given movement using video technology. The position of each joint would be awarded a score, and from this the sum of each of the scores is calculated to determine how risky a particular pose is in relation to moving an object of a certain size or dimensions.
The following video shows the app in action:
The app was tested out based on assessing seventeen different activities that were deemed to be common in warehouses or factories. These were captured using a Microsoft Kinect camera, to create three-dimensional images. In trials the app has been shown to issue warnings for moderately and high risky actions.
The new machine learning approach has been reported to the journal IEEE Robotics and Automation Letters. The paper is titled "Toward Ergonomic Risk Prediction via Segmentation of Indoor Object Manipulation Actions Using Spatiotemporal Convolutional Networks."
More about ergonomics, machine learning, Factory
More news from
Latest News
Top News