A recent study by Massachusetts Institute of Technology researchers shows a new training method for robots may bring them closer to performing complex tasks on job sites with more efficient results, according to Construction Dive.
MIT’s Improbable AI Lab, a group within its Computer Science and Artificial Intelligence Laboratory, combined three types of instruction—language-, visual- and action-based—to teach robots how to perform multistep tasks with higher success rates than comparable training methods.
The training framework, known as Compositional Foundation Models for Hierarchical Planning, reportedly can help robots have an effect in construction, manufacturing and household chores.
The training involves robots approaching a larger objective by completing many smaller steps first. Each step is improved on by the next through a process that allows the training framework to reason about its ideas and take in feedback at each stage to generate a more practical outline.
Anurag Ajay, a Ph.D. student at MIT and a Computer Science and Artificial Intelligence Laboratory affiliate, said robots currently are limited by their hardware but would be good at performing brute force work, such as picking up heavy objects and moving them.
Robots are not yet common on job sites, but as technology improves, contractors are watching the trend of combining AI and robotics for use in construction.
