We evaluated variable autonomy, i.e., who controls the shift between the levels of automation (LoA) for a teleoperated robot navigation task in a simulated office environment. Two variable autonomy control modes were defined-Human-Initiated (HI) and Robot-Initiated (RI) to examine if operators' cognitive workload, situational awareness (SA), and task performance vary between modes. Two levels of automation were used: remotely monitoring (high LoA) and teleoperating (low LoA) the robot while attending to secondary tasks. Task performance metrics included task completion time, collision rates, time spent in the low LoA, and secondary task performance. Operatorrelated measures included workload (NASA-TLX), trust, and perceived situation awareness (SA). Results indicate that the RI control mode outperformed HI in efficiency and reduced workload but compromised operators' SA. Conversely, the HI mode enhanced operator awareness, highlighting a tradeoff between efficiency and situational engagement. This work contributes to understanding the role of variable autonomy in human-robot collaboration and emphasizes the importance of evaluating operator performance in addition to task performance for designing effective teleoperation systems in habituated environments.