Automation: An Engineer’s Perspective

I am a PhD student working with Prof. Julie Shah in the Interactive Robotics Group. A common thread in our research is envisioning a team of humans and autonomous agents accomplishing more together than either can do alone. The applications we consider vary from physical robots working with humans on factory floors, to software agents that monitor team meetings and provide support to help people make better decisions. Working at IRG has given me unique opportunities to observe the interactions between decision-makers for automation, the workers who are the incumbent experts at the task being automated, and designers of the automation systems.

As researchers working on autonomous systems and robotics, it is often too easy for us to think only about expanding the technological frontiers of automation. Traditionally, we have concentrated more on what processes we can automate and less on what are the benefits of automation to all the stakeholders. What makes technological challenges interesting research questions for us also makes them expensive stumbling blocks for companies wanting to automate parts of their processes. Consider automating the task of installing the electrical wiring harnesses in cars during their final assembly. This is a particularly interesting and challenging task for robotics researchers. Designing such a system would involve developing solutions to sense the position of the cable with respect to the robot, reason about how the shape of the cable would change as the robot grasps it and moves it around, and ensuring that the cable is never stretched during the course of the manipulation.

Generally speaking, the arguments presented to automate operations can be classified into direct benefits and indirect benefits. For our task of installation of cables, the direct benefits would include replacing expensive labor hours with potentially less expensive robot hours for the same task. Another benefit might be the increased throughput, if the robot was able to accomplish the installation faster than human workers. With high-fidelity sensing, the installation might also be more consistent with fewer mistakes in adjusting the length of the cables between clamping points. There are also benefits in terms of improving the ergonomics for the assembly line workers by automating installation of wires in areas that are hard to reach.

However, as I said earlier, this task is challenging for even state-of-the-art robotic systems. Newer algorithms and hardware might be needed. An organization that wants to automate this process must estimate what the cost of developing these advancements would be. In addition to the minimum cost of development of new technologies, there are additional sources of cost from ancillary factors. Automating a particular part of the assembly process might impose additional constraints on processes preceding and succeeding it on the assembly line. For example, installation of parts of the wiring harness that cannot be installed by robots might be allocated to the next station. Introduction of automation might necessitate changes to the object to make it more suitable for robotic manipulation. For example, there might be changes to the design of the cables to make them easier for the robot to sense and grasp. Thus the cost of deploying a robotic system might be much greater than the cost of simply developing the enabling technologies—and it is important for a company to strike a right balance between them.

Finally, there is the impact of introduction of automation on the workforce of the company itself. The company might require fewer assembly line workers, but it might now require additional technicians to monitor the performance of the robotic system—to detect its errors and to take corrective action. The company also needs to fully account for all the tasks that the workers who worked on wiring harness installation were performing and reallocate those responsibilities. In cases where automation cannot handle all the tasks, the company must restructure the task allocation among its workers. Failure to do this can result in high error rates on the assembly lines, along with cost and timeline overruns to fix the lapses.

An interaction between research interests, institutions, and workforce make introduction of new automation technologies challenging. One of the biggest hurdles would be incorrect estimation of the benefits, cost, or impact on the current workflow. A tighter integration between the decision-making process to automate, the development process for introducing automation, and the insights of workers would lead to better outcomes for all stakeholders.