In order to explore the influence of the number of supervised vehicles on operators’ workload in the remote supervision of autonomous vehicles, a remote supervision experimental platform is established. A total of 3 to 9 supervised vehicles are set as the independent experimental condition. 40 participants with driving experience and intelligent driving assistance system operation experience were recruited to complete multi-vehicle supervision tasks. Four typical heart rate variability (HRV) indicators, including RMSSD, SDNN, LF, and LF/HF, were collected in the experiment. The NASA-TLX scale was used to measure participants’ subjective task workload. The results show that RMSSD and SDNN decrease overall and maintain a low stable level with the increasing number of supervised vehicles. LF and LF/HF present a trend of initial increase and subsequent decrease. Staged changes are observed in the autonomic nervous regulation responses of participants. All dimensional scores and the weighted task load index (WTLX) of NASA-TLX show an overall upward trend. More significant variations are found in mental demand and temporal demand dimensions. Further difference analysis shows that a relatively stable correlation is maintained between WTLX and the two HRV time-domain indicators (RMSSD and SDNN). Obvious heterogeneous differentiation is identified under the supervision conditions of 8 and 9 vehicles. Continuous accumulation of subjective stress is found in participants under high supervision load. The frequency-domain physiological regulation of HRV is relatively weakened under high supervision load. It is concluded that 3 to 5 vehicles are applicable as the reference range for conventional remote supervision. The vehicle range of 6 to 7 vehicles is defined as a critical interval for further verification and careful evaluation. 8 or more supervised vehicles are not recommended for routine remote supervision without sufficient verification and effective auxiliary measures. This study provides an experimental basis for human-vehicle configuration, workload evaluation and safe operation boundary definition of remote supervision systems for autonomous vehicles.