In large-scale monitoring region

randomly deployed wireless sensor networks may not be fully connected with high probability.Using mobile sink for data collection is one of the feasible solutions.Mobile sink shortest routing prob-lem can be regarded as a special case of TSP with neighborhoods（TSPN） problem

since the neighborhoods are the radio ranges of the sensor nodes

which can be modeled as possibly overlapped disks with diverse sizes.This kind of TSPN problem has no polynomial algorithms so far.To handle it

a novel approximation algorithm was proposed

which first forms a "racetrack" by utilizing the non-intersecting loop property of TSP routes

and then through the inner lane heuris-tic

the bend heuristic and the shortcut searching

the algorithm can find an approximation solution within O（n2） computa-tion time.The formal proofs and the large-scale simulations all verify that our algorithm can achieve a good approxima-tion ratio and can be more efficient than the related algorithms.