Interactive optimizer
Production scheduling
In a flexible manufacturing system, many jobs compete for a handful of machines. Each job is a chain of operations that must run in order and hop from machine to machine along its own route; each machine can only do one operation at a time, and every so often a machine goes down for maintenance. The goal is to sequence everything so the whole batch finishes as early as possible — the makespan.
This is NP-hard: the number of possible schedules explodes with the jobs and machines. An artificial immune system searches it — cloning and hypermutating the best operation-orderings each generation and keeping the ones that pack the Gantt chart tightest. Watch the dashed makespan line pull left as the schedule compacts around the maintenance blocks.
This interactive is a faithful re-design of the research, simplified for illustrative and teaching purposes. It conveys the idea and behaviour of the model — it is not the paper’s full method, data, or results.
Each row is a machine, time runs left to right. Coloured bars are operations (labelled by job); hatched grey blocks are machine maintenance the schedule must work around. The dashed line is the makespan — watch the optimizer pull it left as it reorders operations.
How the model works
Each candidate solution is an operation list — a sequence in which each job appears once per machine, so its k-th appearance is that job’s k-th operation. A decoder walks the list and places every operation at the earliest time its machine is free and its own previous operation has finished, sliding it past any maintenance window that would clash. Because operations are read in job order, job precedence is always respected, and no machine ever runs two operations at once. The makespan is simply when the last operation ends.
The immune optimizer treats good operation lists as high-affinity antibodies: each generation it clones the best ones, hypermutates the clones by swapping operations (more aggressively for weaker solutions), evaluates the offspring, and keeps the fittest — with a trickle of fresh random antibodies for diversity. Over generations the schedule tightens and the makespan drops. This mirrors the immune, harmony-search, and chemical-reaction metaheuristics in the papers below, applied to distributed FMS scheduling subject to machine maintenance.
The research behind it
This demo is a teaching stand-in for a body of work on distributed production scheduling in flexible manufacturing systems, subject to machine maintenance. The full methods, benchmarks, and results are in these papers.
See also the assembly-line-balancing optimizer and the full demo gallery.