This paper deals with the description and testing of the numerical methods employed in the construction of a new multidimensional, explicit, radiation hydrodynamic code for studying protostellar collapse and fragmentation. The code is written in Eulerian spherical coordinates, and the determining equations are solved using second-order-accurate, finite-difference (FD) methods on a radially moving, nonstaggered grid. The set of hydrodynamic equations is advanced using a multistep solution procedure along with either a first-order or a second-order time-integration scheme. On a fixed grid, the code is effectively second-order accurate as demonstrated through convergence testing. For advection on a moving orthogonal mesh, however, the use of standard FD methods results in systematic errors that progressively affect the accuracy of the solution. In order to minimize these errors, a new FD formulation has been derived for both the temporally first-order and second-order versions of the code that yields solutions for the pressureless collapse test case that are numerically invariant under grid motion. Results are also presented for a number of isothermal and nonisothermal protostellar collapse tests that validate this conclusion. Within this new formulation, the radial dependence of the fluid variables is exactly preserved near the origin of the coordinate system. This aspect results in improved local conservation of angular momentum during axisymmetric collapse and in a significant decrease of the level of nonaxisymmetric noise as compared with a previous version of the code. The reduction of systematic sources of error in the new schemes should then allow for a more precise evaluation of the importance of fragmentation during protostellar collapse.