Hierarchical multi-agent predictive maintenance scheduling for trains using price-based approach

While the progress of predictive maintenance has been rising in various application fields and several feasibility studies prototypes have developed, operational implementation requires a more flexible adaptive scheduling predicted interventions. Contrary to traditional preventive tasks that are typically known long advance, may require very short-term schedule changes which also affecting operation. Railway rolling stock some special requirements terms time table adherence vehicle routing make particularly challenging. In this paper, we propose hierarchical multi-agent framework for passenger railway wagons. Besides fulfilling requirement wagon before it fails (and potentially causes delays), demand must be satisfied trains accordingly assigned different routes following table. We distributed learning algorithm using dual decomposition mechanism design approach. The proposed enables preserve local preferences particularities avoid high computational cost. method, first, decompose centralized problem handle fulfillment constraints by Lagrange multiplier (“shadow price”). Furthermore, wagons with private information on system health need perform their failure (with remaining useful lifetime (RUL) corresponding uncertainty). To achieve aim, approach align wagons’ objective function aim central an incentive signal. signal creates non-cooperative game among prove Nash Equilibrium (NE) is optimal solution scheduling. effectiveness method demonstrated case study small network.