Particle-based CFD methods have become very popular in recent years due to their simplified model setup process and ability to free surface problems such as splashing. One of these methods is called Moving Particle Simulation (MPS) and is frequently used in the automotive industry to evaluate splash patterns and churning losses in gearboxes. The method makes an incompressible (constant density) assumption about the fluid which allows it to take larger timesteps than most other particle methods. Historically the disadvantage to particle methods is directly related to their ease of use. The end user is able to preprocess the model by only specifying a fixed particle size which the software uses to discretize the model, providing constant spatial resolution of the domain. This particle size must be set based on the smallest feature so large models with small features require a large number of particles. As the number of particles more memory and computational capacity are required which can make problems impractical to solve. A variety of approaches to handle this issue have been developed, each with its own advantages and disadvantages. The fundamental challenge is that particle methods are able to guarantee mass and momentum conservation by virtue of resolving fluid with discrete and persistent particles, but this guarantee must be relaxed to implement multiple particle sizes in a single model. The MPS approach allows one or more regions to be defined where the particle size is some fraction of the global size. Around these regions a transition zone is made where particles are allowed to subdivide (or merge) such that mass and momentum and conserved as well as possible. This paper will discuss the challenges inherent to variable-resolution particle methods, some existing methods, and the new approach taken by MPS.