Managed lanes (MLs) system has been widely implemented on freeways to mitigate congestion and improve efficiency. MLs are usually designed as concurrent with general-purpose lanes (GPLs), and separated by a barrier or painted stripe with several at-grade ingresses and egresses. However, these kinds of designs may result in weaving segments between ingress (egress) and on-ramp (off-ramp). For example, vehicles from on-ramp must cross multiple GPLs to get access to the ML. This study aims to investigate the safety effects of weaving length, traffic condition, and driver characteristics on drivers’ mandatory lane change behavior based on a driving simulator study. Mixed factorial design with two within-subject factors (traffic volume: off-peak and peak; speed harmonization (SH): SH and Non-SH) and one between-subject factor (weaving length per lane change (LLC): 600 feet, 1000 feet, and 1400 feet) were employed in this study. Fifty-four licensed drivers were recruited to conduct this driving simulator experiment. Based on the experimental data, three lane change decision metrics (i.e., lane change merging gap, duration, and patience time), three lane change execution metrics (i.e., maximum longitudinal deceleration, lateral acceleration, and steering wheel angle), and two surrogate-safety metrics (i.e., number of conflicts and time exposed time-to-collision) were analyzed. Results indicated that for the ingress of MLs (entrance weaving segment), 1000 feet LLC would be recommended if the space is limited, otherwise 1400 feet LLC is preferable. For the egress of MLs (exit weaving segment), however, only 1000 feet LLC was recommended since 1400 feet LLC was found to be significantly more dangerous than 600 and 1000 feet LLC. Moreover, peak traffic condition could significantly increase the difficulty of lane change behavior on the weaving segments, and speed harmonization could significantly improve the lane change safety on the entrance weaving segments.