Tartarus is a standard test problem used to evaluate evolutionary computation techniques for solving problems in artificial intelligence. A gap in the Tartarus literature is a lack of systematic baseline studies for standard types of chromosomes in broad use in evolutionary computation. In this paper we adapt plain finite state automata to serve as controllers for virtual robots in the Tartarus environment. We overcome the bandwidth limitations on finite state automata that have prevented their use in Tartarus thus far by permitting a finite state machine to simultaneously generate a Tartarus action and select which of eight sensors will supply its next input. We show by simulation that our finite state chromosome outperforms published representations without internal state information but is outperformed by some chromosomes that use internal state information as part of a more complex structure. A summary of various technologies used thus far for the Tartarus problem and their best results is given.