A Multi-agent System Embedding Online Tensor Learning for flow shopScheduling

In this study, we introduce a multi-agent system for flow shop
scheduling with a novel online tensor learning component which
identifies recurring patterns of elements and shares this information
between agents to guide the search process.

Each agent instantiates the same heuristic and starts the search in
parallel from a different point in the space for solving the given flow
shop scheduling problem instance. The inter-agent communication protocol
enables collection of incumbent solutions forming a 3 dimensional (3rd
order) tensor. The resultant matrix after factorisation of the tensor is
used by each agent as a seed to form a starting solution from which the
search process continues.

The tensor based multi-agent approach is evaluated using well-known flow
shop scheduling benchmarks. The results show that the use of tensor
analysis improves the overall performance of the approach when compared
to a variant using another technique. Moreover, the proposed approach
proves to be promising outperforming standard heuristics and matching
overall performance when compared to the state-of-the-art.