A new approach to the problem of attitude determination using the global positioning system, which parameterizes the rows of the attitude matrix, differing itself from other methods (i.e., the method of parallel to Wahba's problem (PWP) or the traditional linearized method), is presented. The new approach is a forward procedure and does not need any initial attitude knowledge. Thus, it is more computationally economical than the linearized method, which employs an iterative procedure. Unlike either the linearized method or the PWP method, which require at least two baselines, the approach here is suitable for any baseline configurations, including not only either coplanar or noncoplanar baseline configurations, but also one-baseline configurations. If the antennas are carefully arranged to satisfy the symmetric condition, the derived solution can be further simplified. The concept of the symmetric condition is applicable for any baseline configurations, whereas the concept of the balanced condition proposed in the PWP method is limited to the case of noncoplanar baseline configurations. Both concepts are identical in the noncoplanar case. Therefore, the symmetric condition can be regarded as a natural generalization of the balanced condition. The results of experiments demonstrate that algorithms derived from this new approach are highly efficient.