Bus-Conditioned Zero-Shot Trajectory Generation via Task Arithmetic

Shuai Liu, Ning Cao, Yile Chen, Yue Jiang, Gao Cong.

“Bus-Conditioned Zero-Shot Trajectory Generation via Task Arithmetic” proposes MobTA, a framework for generating mobility trajectories in a target city without access to its real mobility data. Using source-city mobility trajectories and publicly available bus timetables from both cities, MobTA learns the parameter shift from bus-timetable-based generation to mobility trajectory generation in the source city and transfers it to the target city through task arithmetic. Experiments demonstrate that MobTA outperforms existing methods and approaches the performance of models fine-tuned on target-city mobility data.