For　shape　optimization　of　truss　structures,　the　design　variables　are　separated　into　two　kinds　of　bar　sectional　size　and　nodal　position　by　multi-level　decomposition　optimization　method.　The　solution　process　is　divided　into　two　steps.　At　the　first　step,　supposing　all　nodal　positions　are　unchangeable.　The　bar　sectional　sizes　are　optimized　to　make　the　weight　of　the　structure　minimized　under　constraints　of　stress,　displacement　and　local　stability.　At　the　second　step,　supposing　the　active　displacement　constraints　of　the　first　step　keeping　unchanged.　A　quadratic　programming　model　that　increases　the　structural　rigidity　is　solved.　The　obtained　new　nodal　positions　subject　to　the　constraints　and　make　the　value　of　the　objective　function　equal　to　or　less　than　that　of　the　first　step.　The　two　steps　alternatively　executed　until　convergence　comes　off.