Inertial-based　upper　limb　motion　capture　technology　is　widely　applied　in　directly　controlling　robots　through　human　motions　due　to　its　low　cost　and　portability.　Most　upper　limb　motion　capture　models　typically　focus　solely　on　estimating　the　wrist　position　(WP)　while　neglecting　the　motion　of　the　wrist.　Moreover,　these　models　struggle　to　precisely　estimate　joint　positions,　especially　in　the　presence　of　magnetic　disturbances.　In　this　article,　a　two-layer　upper　limb　motion　capture　model　is　proposed.　In　the　upper　layer　of　the　model,　a　hand　center　position　estimation　method　(HCPEM)　is　presented　to　estimate　accurately　hand　center　position　(HCP)　regardless　of　magnetic　field　interference.　In　the　lower　layer　of　the　model,　an　optimized　sensor-to-segment　alignment　algorithm　based　on　physical　constraints　(OABPCs)　is　proposed　to　align　sensors　and　segments　with　the　aim　of　improving　the　precision　of　the　estimated　HCPs.　The　UR3　robotic　experiment　platform　was　designed　to　verify　the　accuracy　of　the　HCPs.　For　different　motion　tasks　and　different　mounting　positions　of　the　inertial　measurement　units　(IMUs),　the　performance　of　the　proposed　two-layer　model　was　further　investigated　through　experiments.　The　results　show　that　the　HCP　estimated　by　the　proposed　two-layer　model　achieves　root-mean-square　errors　(RMSEs)　below　2.5　cm　for　different　motion　tasks　and　various　IMU　mounting　positions,　even　in　the　presence　of　magnetic　field　disturbances.　Compared　with　the　existing　algorithm,　the　RMSE　of　the　joint　position　measurements　is　improved　by　1　cm.　This　study　enables　flexible,　accurate,　and　convenient　control　of　the　robot　through　human　motion.