A　method　to　identify　the　dynamic　model　parameters　of　modular　robot　with　flexible　joint　using　3D　motion　measurement　system　Optotrak3020　is　proposed.　Firstly,　the　robot　dynamics　model　is　parameterized　in　a　linear　form　to　exclude　the　stiffness　torque　and　the　scaling　problem　is　eliminated　from　the　parameter　matrix.　The　exciting　trajectory　is　based　on　the　finite　Fourier　series　function　and　the　coefficients　of　the　Fourier　series　are　optimized　with　the　self-adaptive　genetic　algorithm.　When　each　joint　of　robot　is　commanded　to　follow　the　single-joint　optimized　exciting　trajectory,　the　motor　position,　link　position　and　motor　torque　are　measured　and　recorded　in　real　time.　Finally,　using　the　classic　linear　least　square　parameter　identification,　the　dynamic　parameter　identification　experiment　is　carried　out　on　the　first　three　joints　of　PowerCube　modular　robot,　and　parameters　of　robot　dynamic　model　are　derived.