Minimizing　the　mass　and　moment　of　inertia　is　a　crucial　objective　in　robot　design.　Especially　for　the　biped　humanoid　robots,　leg　mass　and　moment　of　inertia　severely　affect　the　robot＇s　ultimate　speed,　motion　stability,　and　interaction　safety.　The　lightweight　design　for　the　legs　of　biped　humanoid　robots　has　been　a　hot　but　difficult　research　topic　in　recent　decades.　This　paper　will　propose　a　new　optimization　approach　for　achieving　lightweight　design　of　biped　humanoid　robot　legs.　Firstly,　the　joint　drivetrain　dynamic　model　will　be　established,　followed　by　the　process　of　determining　the　selection　criteria　of　motors　and　gearbox,　and　clarifying　the　calculation　method　of　joint　mass　attributes.　Secondly,　the　minimal　total　mass　of　the　robot＇s　legs　is　taken　as　the　goal　to　optimize,　and　the　parameters　related　to　the　motor　and　gearbox　models　are　the　design　variables.　As　the　robot　walks　stably,　the　maximal　walking　speed　that　is　close　to　the　target　speed　is　regarded　as　the　constraint.　A　complex　method　is　then　implemented　in　a　commercial　mathematical　software,　the　model　to　simulate　robot　dynamics　is　established　in　commercial　dynamic　software,　and　the　dynamic　simulation　is　completed　using　the　three-dimensional　linear　inverted　pendulum　gait　planning　method.　Finally,　the　Walker　robot　is　used　as　an　example　to　demonstrate　the　effectiveness　of　the　proposed　design　optimization　approach.　The　results　show　that　the　design　optimization　method　can　significantly　reduce　the　total　mass　of　the　robot＇s　legs,　reduce　the　torque　requirements　of　the　robot＇s　leg　joints,　and　improve　the　stability　of　the　robot＇s　motion.　The　optimization　approach　presented　in　this　study　is　also　important　and　applicable　to　the　lightweight　design　of　other　categories　of　robots.