This paper presents a path planning algorithm to reduce localization error by intelligently choosing a path that will result in low expected sensor errors. For example, Mars rovers can enhance localization accuracy by selectively driving over feature-rich terrain where visual odometry can be used. In general, having an accurate localization is vital for autonomous mobile exploration platforms such as rovers and aerial vehicles. However, typical path planning methods tend to narrowly focus on minimizing path length. Our proposed path planning algorithm, namely Error Propagation A∗ (EPA∗ ), intelligently balances path length and localization. EPA∗ is a graph search algorithm, where a linear error propagation law is derived before the search on each edge of the graph. Using the propagation law, EPA∗ can quickly find a path that minimizes a given objective function, which includes both path length and error covariance. We demonstrate the EPA∗ algorithm using the real data from Curiosity. The result demonstrates that the EPA∗ algorithm can find a path that balances the path length and the expected localization error, as expected.