Abstract: Path planning plays an important role in robot-assisted flexible needle insertion procedures. The trajectory errors derived from the needle-tissue interaction can be predicted based on a proper needle-tissue interaction model, which is important for needle path planning, especially when the needle is inserted into highly deformable tissues. A novel iterative learning-based path planning method is proposed. First, a set of candidate paths are offline generated according to the reachable domain analysis under non-deformable environment, wherein the best candidate path is selected based on genetic simulated annealing algorithm (GSAA). Then, a real-time needle-tissue interaction model is utilized to predict the target deformation and needle deflection, which is combined with an iterative learning control-based path correction algorithm to achieve satisfactory trajectory accuracy. Finally, a global sensitivity analysis is applied to the proposed planning framework for parameter optimization. Needle insertion experiments are conducted with pre-defined targets and obstacles in phantoms to verify the proposed path planning methods. The results show that the error of the proposed algorithm can reach an average error below 0.45 mm.

Key words: iterative learning control, needle trajectory planning, minimally invasive needle insertion, flexible needle