In this paper, two different three-dimensional (3D) indoor visible light positioning (VLP) algorithms are experimentally assessed for an industrial environment. The Cayley–Menger determinant (CMD) and linear least square (LLS) trilateration algorithms use the received signal strength (RSS) to estimate the receiver’s 3D position without prior knowledge of its height. The unknown 3D position of the receiver is estimated by the trilateration algorithms coupled with a cost function under different realistic scenarios. The performances of the algorithms are experimentally evaluated in terms of positioning error by considering two different light-emitting diode (LED) configurations in the presence of different receiver tilt angles, and with multipath reflections. It is observed that the widespread square LED configuration results in position ambiguities while a star-shaped configuration is much more accurate. Experimental tests performed in a 4 m 4 m 4.1 m area with four LEDs reported a median positioning error of 10.6 and 10.5 cm using the LLS and CMD algorithms, respectively, without the presence of receiver tilt or multipath reflections. However, when a receiver tilt of 10 was added, the median error increased to 22.7 cm using the LLS algorithm and 21.6 cm using the CMD algorithm. Overall, the achieved mean and maximum values using the LLS algorithm were 13.1 and 39 cm, respectively, while they were 12.2 and 34 cm using the CMD algorithm.