Spacecraft operation in proximity to small bodies requires advanced guidance, navigation and control (GN&C) protocols that are able to operate and land autonomously. To achieve a successful landing, the on-board GN&C algorithm must reliably land the spacecraft near the targeted surface location, while ensuring a low velocity at touchdown. In this paper, the soft landing problem is reformulated as a convex optimization problem, and Model Predictive Control (MPC) is used to handle both the optimization process and system constraints in a unified framework. We present a new collision avoidance method; an optimal separating hyperplane coupled with a projection theorem argument to define auxiliary set points used in the MPC formulation. We show that these auxiliary set points converge to the desired target state, and thus the spacecraft will reach its goal safely. Numerical simulation results demonstrate the performance.