The　existing　kinematic　synthesis　of　serial　robots　generally　focuses　on　configurations　with　specified　degrees-of-freedom　(DOFs).　This　paper　proposes　a　new　generation　method　of　configurations　based　on　iteration　concepts,　defines　the　concept　of　a　configuration　tree,　and　establishes　the　tree-evolution　kinematic　synthesis　framework.　The　configuration　tree　includes　configurations　with　different　DOFs,　and　the　‘tree’　reflects　the　configuration　relationship.　The　objectives　of　this　study　are　to　research　the　performance　distribution　characteristics　of　the　‘tree’,　develop　the　tree-evolution　strategies,　and　propose　a　serial　robot　configuration　model　with　hyperparameters.　The　study　observed　that　many　typical　performances　have　a　monotonic　distribution　on　the　‘tree’.　Based　on　the　estimation　concept,　the　tree-evolution　strategies　search　for　the　required　configurations　with　high　efficiency.　In　the　kinematic　synthesis　task　of　an　eight-DOF　robot,　the　efficiency　of　configuration　classification　is　increased　by　at　least　30%,　and　the　efficiency　of　configuration　optimisation　is　increased　by　65%.　Moreover,　the　calculation　efficiency　is　improved　further　as　the　DOF　increases.　The　example　shows　that　the　new　method　improves　the　calculation　efficiency　of　the　process　of　kinematic　synthesis　compared　with　that　for　the　traditional　method.　©　2021　Elsevier　Ltd