In order to solve the problem of invalid segments in collected GNSS (global navigation satellite system) paths of wheeled mobile robots, an invalid path elimination method based on self-adaptive segmentation and rearrangement is proposed, and a path smoothing method based on multi-objective optimization is presented. Firstly, the collected GNSS paths are divided into DRD (drive-reverse-drive) segments by points' heading, and some elimination rules are set to eliminate the invalid segments. For the discrete segments after eliminating invalid segments, they are divided for the second time according to the length of adjacent segments, so as to achieve rapid rearrangement of effective segments. Then a multiobjective optimization function, asymptotic boundary point constraints and rectangular region constraints are established with the rearranged point sets as decision variables, and the multi-objective optimization problem is transformed into quadratic programming for optimal solution, in order to reduce the position difference with the original path while improving the path smoothness. Experimental results show that the proposed method can effectively deal with invalid GNSS paths collected in different road forms, the average position difference between the optimized results and rearranged path is less than 0.2 m, the average processing time is 8.8 ms, and the processed path can be applied to trajectory following of unmanned vehicle.