Legible Lights for Factory Robots

How can simple lights be used in low DOF factory robots to send legible signals to factory workers?

What can we learn from robotics company employees about important factors when designing robot communications?

Key Takeaways: Employees who work with robots have valuable insights into how and when robots should communicate, and these communications may require different signal styles for different communication needs.

Industry & Academia

In this collaborative work, we leveraged employee knowledge and experiences to identify opportunities for improving the communication capabilities of the robots, resulting in the addition of several robot state communications to their initial software set leveraging both employee- and social robotics literature- sourced ideas for communicating with lights.

OTTO 100 Robots

The OTTO 100 robots (made by OTTO Motors) are small, rectangular box shaped self-driving vehicles that can carry up to 100kg. They are used in factories to transport products and supplies.

They have one visual light channel, one sound channel, and two degrees of freedom in motion

How do we know what to communicate?

A company-wide survey was developed to extract needs and ideas for robot expressive communications from the people who best understand the broad range of contexts and the end users that utilize the robots.

Why survey the Clearpath employees? The surveyed employees have a wide spectrum of specialties, from customer support, to robot testing, to building the robots, which may give them unique insights as to where it might be valuable to improve the robots.

New States and Lights

Base on the survey results, new communication states and lights were added to the robot software and tested in a user study.

Two new communication states were added: (1) blocked and (2) idle

Three new light "personas" were tested: (1) car-like lights, which mimics how a vehicle signals, (2) sweeping lights, which move from the back of the robot around to the front, and (3) heartbeat lights, which pulse in a solid color.

Car-like lights are similar to car blinkers for the "turn" state

Sweeping lights at the goal move around the robot from center back to center front in the "at goal" state

Heartbeat lights pulse in red in the "blocked" state

Software Flow

First, the user externally set which light "persona" they desired.

As the robot goes about its tasks, it uses information from its sensors to determine which state it is in. If one of the states needing a light pattern is triggered, the robots will display the relevant state light pattern until that state is exited.

To avoid having to create many PNGs to experiment with the lights, a dynamic light layer was written using C++ that listened to a ROS topic of color values, wrote those values to a row of a PNG, and used that PNG to send to the LED strip. Different python scripts were created for each light set that published to the ROS topic. The python scripts controlled the color, spacing, and timing of the lights and when different patterns were activated.

User Study

The goal of this study was to do an initial evaluation of the relative merits of the four light "personas."

The backdrop of the study took inspiration from a common problem at a specific customer site involving an L-junction and several moving entities backing in and out of the space. In the customer site, forklifts would occasionally block the OTTO 100s, and because of their respective dropoff locations, the OTTO 100s would occasionally block the forklift drivers also. This scenario was chosen because understanding robot state well in an area with occlusions and the potential for multiple actors would be critical for rapid and successful coordination.

Overall, all participants preferred the custom lights with additional states over the pre-existing lights.

Sweeping lights were the clearest for the "blocked" state because they were flashy and attention-grabbing.
Heartbeat lights were preferred for the "at goal" state because they are visible from any angle, which is ideal for directionless states like "at goal."
Car-like lights were preferred for the "turn" state because they were familiar signals used in driving.

Key Takeaways

People making and working with robots have valuable insights about their communications.
The participatory approach of the company survey surfaced desired robot state communications that were needed, but did not previously exist.
Familiarity of the robots, or working on them directly, may not predict how people view light communications.
In different situations, different styles of light communication are preferred.
All of the new light sets outperformed the factory default

Publications:

Bacula, Alexandra, Jason Mercer, Jaden Berger, Julie A. Adams, Heather Knight. “Integrating Robot Manufacturer Perspectives into Legible Factory Robot Light Communications,” Transactions on Human-Robot Interaction. Accepted October, 2022.
Bacula, Alexandra, Jason Mercer, and Heather Knight. "Legible Light Communications for Factory Robots." In Companion of the 2020 ACM/IEEE International Conference on Human-Robot Interaction, pp. 119-121. 2020.

Page updated

Report abuse