A dynamic heuristic for the real time hoist scheduling problem

Considers a multi-product production (electroplating line) in a real time context :jobs arrive randomly at the input buffer. Then, any time a new job arrives at the input buffer, given the state of the whole line (the position and the remaining soaktime of all the already in process products) and given products at the input buffer a new schedule of the whole jobs must be computed. The authors refer to this problem that must be solved successively as the dynamic hoist scheduling problem (DHSP). Usually, in order to solve the DHSP, a heuristic is introduced. It assumes that the sequencing of the already in process jobs does not change. Then, a few seconds computation gives a feasible solution. But it appears that the production of the line might be improved with a better scheduling strategy. The authors´ approach consists in improving the computation time per DHSP (in order to respect real time computation limits) while preserving the production efficiency. So, a branch and bound search is still used in order to optimise their criterion. But, the authors propose to replace the classical backtrack algorithm with the Ginsberg dynamic backtracking algorithm. Finally, the authors introduce a bound dependent dynamic heuristic based on the use of dynamic backtracking