An effective mirror-based genetic algorithm for scheduling multi-mode resource constrained projects
RIS ID
131967
Abstract
This paper presents an effective mirror-based genetic algorithm for scheduling resource-constrained multimode projects. Towards balancing the exploration of search space versus the exploitation of high quality solutions, the algorithm uses an array of synergetic techniques, and a feature which allows inverting the problem and solving the inverted problem. With this feature, the algorithm solves both the original and the inverted problems independently and among the obtained solutions chooses the best. The algorithm has been tested on a large number of various benchmark instances, and the results of computational experiments indicate that it can find the best available solutions in the literature for 548 out of 552 PSPLIB instances. For each of the 4 instances, 552-548, which it has missed to find their best available solutions in the literature, the difference of the obtained solution from the best available solution is only 1 unit. The procedure has also been tested on 4320 extra benchmark instances of MMLIB50, MMLIB100, and MMLIB+. With the limit of 50,000 schedules on the set MMLIB+, which includes 3240 instances, the procedure outperforms all other procedures reported in Van Peteghem and Vanhoucke (2014).
Publication Details
Zamani, R. (2019). An effective mirror-based genetic algorithm for scheduling multi-mode resource constrained projects. Computers and Industrial Engineering, 127 914-924.