Graph Learning Based solution for MRTA

Author: Steve Paul

Efficient multi-robot task allocation (MRTA) is fundamental to various time-sensitive applications such as disaster response, warehouse operations, and construction. MRTA involves using a team of robots to perform tasks that are spatially distributed, and present time deadlines and different workloads that may require multiple visits by robots to complete the task.

Here, we proposed encoder-decoder graph neural network involving Capsule networks and multi-head attention mechanism, and innovatively add topological descriptors (TD) as new features to improve transferability to unseen problems of similar and larger size. Persistent homology is used to derive the TD, and proximal policy optimization is used to train our TD-augmented graph neural network. The resulting policy model compares favorably to state-of-the-art non-learning baselines while being much faster. Our solution was the runner up in 2022 ICRA outstanding coordination paper award.
For more details, check out the papers below.