We present an energy-efficient implementation of RGB-D simultaneous localization and mapping (SLAM) by applying approximate computing (AC) techniques such as loop perforation (LP) and reduced precision (RP). To reduce processing time and power consumption, LP and RP were applied to the two most computationally challenging portion of the multikernel pipeline of SLAM, i.e., SIFT feature detection/extraction and graph-based global pose optimization. We discover that approximation errors on the target kernels can be tolerated due to not only the inherent robustness of the kernel itself but also repairing process of the subsequent functions. Experimental results, assuming the approximated hardware architecture, demonstrate that our approximated SLAM can achieve 48.5% reduction in execution time and up to 71.8% power reduction for the common indoor SLAM scenarios-with negligible changes in quality. Index Terms-SLAM, approximate computing, loop perforation, reduced precision, SIFT, G2O, PCG.