In practice, additional knowledge about the target to be tracked, other than its fundamental dynamics, can often be modeled as a set of soft constraints and utilized in a filtering process to improve the tracking performance. This paper develops a general approach to the modeling of soft inequality constraints, and investigates particle filtering (PF) with soft state constraints for target tracking. We develop two PF algorithms with soft inequality constraints, i.e., a sequential-importance-resampling particle filter and an auxiliary sampling mechanism. The latter probabilistically selects the candidate particles from the soft inequality constraints of the state variables so that they are more likely to comply with the soft constraints. The performances of the proposed algorithms are evaluated using Monte Carlo simulations in a target tracking scenario.