"There's a long history of debate on whether we want to build a single, powerful humanoid robot that can do all the jobs, or we have a team of robots that can collaborate," said Hao Zhang, associate professor at UMass Amherst's Manning College of Information and Computer Sciences and director of the Human-Centered Robotics Lab.
In the context of manufacturing, a team of robots can be more cost-effective by optimizing the strengths of each unit. However, coordinating a diverse set of robots-some stationary, some mobile, with varying capabilities-presents a significant challenge.
Zhang's team addressed this challenge with their LVWS approach. This system allows robots to form subteams and voluntarily wait when necessary, ensuring more efficient task completion. "Robots have big tasks, just like humans," Zhang explained. "For example, they have a large box that cannot be carried by a single robot. The scenario will need multiple robots to collaboratively work on that."
Voluntary waiting plays a key role in this system. "We want the robot to be able to actively wait because, if they just choose a greedy solution to always perform smaller tasks that are immediately available, sometimes the bigger task will never be executed," Zhang added.
The team tested their approach by assigning six robots 18 tasks in a computer simulation, comparing LVWS against four other methods. While the alternative methods showed suboptimality ranging from 11.8% to 23%, the LVWS method achieved a suboptimality of only 0.8%. "So the solution is close to the best possible or theoretical solution," said Williard Jose, a doctoral student in computer science at the Human-Centered Robotics Lab and co-author of the paper.
Jose provided an example to illustrate the advantage of waiting: if two small robots can lift four pounds each, and one is busy while a seven-pound box needs to be moved, it's better for the available small robot to wait for the other, rather than have a larger robot-better suited for other tasks-handle the box alone.
While it might seem logical to always calculate the optimal task allocation, Jose noted the impracticality due to time constraints, especially with a larger number of robots and tasks. "The issue with using that exact solution is to compute that it takes a really long time," he explained.
In scenarios involving 100 tasks, where an exact solution is impractical, the LVWS method completed tasks in 22 timesteps, outperforming comparison models that ranged from 23.05 to 25.85 timesteps.
Zhang envisions this research advancing the development of multi-robot systems, particularly in large-scale industrial settings. "A single humanoid robot may be a better fit in the small footprint of a single-family home, while multi-robot systems are better options for a large industry environment that requires specialized tasks," he said.
Research Report:Learning for Dynamic Subteaming and Voluntary Waiting in Heterogeneous Multi-Robot Collaborative Scheduling
