This study proposes an improved Monte Carlo method, considering that the traditional method lacks precision while calculating the workspace of a robot. The improved Monte Carlo method comprises two stages. In the first stage, a seed workspace is generated using the traditional Monte Carlo method. In the second stage, the seed workspace is expanded based on the normal distribution, and each region in the obtained workspace can be accurately described by setting an accuracy threshold in the process of expansion. Taking into account the characteristics of the normal distribution, to improve the efficiency of the expansion, dynamically adjustable standard deviations are used. Based on the obtained workspace, a voxel algorithm is proposed to determine the volume of the workspace. The algorithm for searching the boundary has been designed to locate the boundary as well as the non-boundary of the workspace. Refining the boundary alone reduces the calculation time and the resulting error. In order to verify the validity and practicability of the algorithm, the improved Monte Carlo method and the proposed volumetric algorithm were simulated and analyzed using a 9-degrees-of-freedom super-redundant serial robot. The results show that when the number of sampling points is the same, the boundary of the workspace generated by the improved Monte Carlo method is smoother and the noise is smaller. When the accurate workspace is obtained, the number of sampling points needed by the improved method is only 4.67% that of the traditional method. The designed volumetric algorithm is also more efficient, with a relative error less than 1%. The volume of workspace thus obtained can be used to evaluate the performance of a serial robot, which lays a theoretical foundation for the subsequent optimization of serial robot configuration.