The　genetic　algorithm　(GA)　for　multi-head　surface　mounting　machines　is　studied,　and　it　has　two　contributions　to　the　multi-head　case,　one　is　that　the　algorithm　has　a　special　crossover　operation;　the　other　is　that　its　mutation　operation　is　composed　of　exchange　operator　and　inversion　operator.　Experimental　results　show　that　the　proposed　approach　gives　good　robot　assembly　time　and　less　movement　of　the　feeder　carrier　and　PCB　table.　©　2008　Springer　Berlin　Heidelberg.