Space　mapping　(SM)　methodology　has　been　recognized　as　a　powerful　tool　for　accelerating　electromagnetic　(EM)-based　yield　optimization.　This　paper　proposes　a　novel　parallel　space-mapping　based　yield-driven　EM　optimization　technique　incorporating　trust　region　algorithm　and　polynomial　chaos　expansion　(PCE).　In　this　technique,　a　novel　trust　region　algorithm　is　proposed　to　increase　the　robustness　of　the　SM　surrogate　in　each　iteration　during　yield　optimization.　The　proposed　algorithm　updates　the　trust　radius　of　each　design　parameter　based　on　the　effectiveness　of　minimizing　the　l(1)　objective　function　using　the　surrogate,　thereby　increasing　the　robustness　of　the　SM　surrogate.　Moreover,　for　the　first　time,　parallel　computation　method　is　incorporated　into　SM-based　yield-driven　design　to　accelerate　the　overall　yield　optimization　process　of　microwave　structures.　The　use　of　parallel　computation　allows　the　surrogate　developed　in　the　proposed　technique　to　be　valid　in　a　larger　neighborhood　than　that　in　standard　SM,　consequently　increasing　the　speed　of　finding　the　optimal　yield　solution　in　SM-based　yield-driven　design.　Lastly,　the　PCE　approach　is　incorporated　into　the　proposed　technique　to　further　speed　up　yield　verification　on　the　fine　model.　Compared　with　the　standard　SM-based　yield　optimization　technique　with　sequential　computation,　the　proposed　technique　achieves　a　higher　yield　increase　with　shorter　CPU　time　by　reducing　the　number　of　SM　iterations.　The　proposed　technique　is　illustrated　by　two　microwave　examples.