Traditional industrial robots perform a series of tasks which are typically segregated from the work that humans do. This means that robot-human workflows are, for the most part, separate, serialized workflows. This situation has evolved mostly because companies don’t want their robots hurting their humans, or their humans interfering with their robots.
Segregation is expensive and inefficient. But, what if you had a way to integrate robot and human workflows? What if you could design processes – and robots – so that robots could hand off to humans seamlessly, and humans could hand off back to robots in the other direction? Or, even more forward-looking: what if we were able to get robots and humans to partner together deeply on some process or task?
The Collaborative Robotics research project explores the challenges of building safe, secure and effective Human-to-Robot workflows. This work is about combining strengths of humans and robots in a seamless way to maximize productivity, safety and security in a continuously changing environment.
In the world of humans, good decision-making is typically aided by good information. To improve decision making by collaborative robots are performing experiments on datasets from different domains, such as health, banking and agriculture. These experiments allow us to test new models for simplifying and optimizing the data that the cobots need to analyze.
For example, to train a model that determines whether a cobot is performing a specified job accurately, requires delivering specific sensor data to the cobot that includes only “good” or non-faulty sensor data versus traditional models that would include all sensor data, including faulty or “bad” sensor data.
Addressing such key challenges as human-to-robot workflows will require new technologies, including machine learning, to build the next generation of robotics solutions that will allow machines to unleash our human potential.