Robots in Warehouses: MIT Research Makes Operations More Efficient
Hundreds of robots are now an integral part of many industries, from e-commerce to automotive production. But managing 800 robots efficiently in a warehouse is no easy feat. That’s where MIT researchers step in with their AI-driven solution.
The team at MIT took inspiration from traffic congestion algorithms to develop a deep-learning model for warehouse management. This model encodes crucial information about the warehouse, robots, paths, tasks, and obstacles to predict the best areas for decongestion.
By dividing the robots into smaller groups, the researchers were able to decongest them faster using traditional coordination algorithms. This approach improved efficiency significantly, making the process nearly four times faster than random search methods.
Apart from warehouse operations, this innovative approach could also be applied to other complex planning tasks like computer chip design. The lead author Zhongxia Yan, a graduate student, will present this work at a prestigious conference.
“Robotic Tetris” in Warehouses Explained
When customer orders come in, robots grab the requested items from shelves and deliver them to human operators for packing. These robots resemble a high-tech game of “Tetris” in a giant warehouse, with the risk of collisions if paths intersect.
Traditional algorithms avoid crashes by replanning trajectories, but with real-time operations in a busy warehouse, speed is crucial. Machine learning focuses on congested areas to reduce travel time effectively.
The researchers built a neural network that segments robots into smaller groups for efficient planning. By considering relationships between robots and minimizing computational repetition, they managed to decongest warehouses up to four times faster than traditional methods.
Future plans include deriving simple, rule-based insights from the neural network for easier implementation in actual warehouse settings. This groundbreaking work was supported by Amazon and the MIT Amazon Science Hub.