In this paper, two algorithms of Global Positioning System based attitude determination are proposed. The first algorithm extends the Kalman filter approach to determine the integer ambiguity and the orientation that is needed in a typical gps-based attitude determination problem. The second algorithm explores the mean field annealing neural network approach, which is a combination of the competitive Hopfield neural network and the stochastic simulated annealing technique, to resolve the optimal attitude problems. A test platform is set up for verifying these algorithms. The two algorithms are further compared in terms of computation speed and convergence rate.