In　the　field　of　obstacle　avoidance　path　planning,　the　traditional　Rapidly-Exploring　Random　Tree　(RRT)　algorithm　has　many　problems,　such　as　no　direction　and　low　efficiency.　So　it　is　often　used　to　adjust　the　growth　direction　of　random　tree　nodes　by　introducing　a　target　bias　strategy　to　decrease　the　search　blindness.　On　this　basis,　the　end　movement　distance　and　the　variation　range　of　each　joint　during　the　manipulator　trajectory　planning　process　have　been　focused　on　in　this　paper.　Considering　the　requirements　of　the　speed　of　the　planning　executable　trajectory　and　the　smoothness　of　the　moving　process,　a　cost　function　about　the　path　length　and　the　smooth　change　of　the　joints　has　been　designed.　Then,　under　the　premise　of　the　stability　of　the　path　planning　results,　an　improved　RRT　algorithm　on　dynamical　adjustment　of　the　new　nodes　generation　has　been　proposed　to　increase　the　planning　efficiency　obviously.　Its　feasibility　and　effectiveness　have　been　verified　fully　by　a　series　of　simulation　experiments　based　on　MATLAB　platform.　©　2021,　Springer　Nature　Switzerland　AG.