Firstly, at crowded stations it is recommended the use of separated doors for boarding and alighting. According to our
results, the Passenger Service Time (PST) could be reduced by 31–82% compared to the use of two-way doors. In addition,
this measure produces less interaction between passengers and therefore lower densities and dissatisfaction.
Second, regarding to the location of vertical handrails, it was found that the best option is the use of a central handrail
located in the middle of the door frame. However, in order to generate a one-way flow through each side of the handrail,
signaling and marking is required. Under these circumstances, the PST could be 13–34% lower compared with the case where
the handrail is in the center of the hall of the car. In addition, to reduce the density on the platform, it is recommended that
the handrail must be displaced about 0.40 m inwards the door frame, so that the passengers coming from the platform do
not perceive the handrail as an obstacle. If the handrail cannot be located in the middle of the door frame, it is advisable to
use two handrails on both sides of the door located approximately 0.80 m from the center of the hall.
Third, in relation to the use of a keep-out zone on the platform, the simulation showed that PST was reduced to 50%. However,
in the experiments this measure slightly increased the PST. This was an experimental mistake because either: (a) the
subjects did not follow the instructions; or (b) the instructions were not clear enough to be followed. Nevertheless, the experience
shows what would happen if the scheme is applied without the appropriate directions and enforcement. However in
the 30% of the laboratory observations in which the keep-out zone worked as expected, passengers formed three lanes for
alighting and two lanes for boarding. As a consequence, less dissatisfaction inside the train was obtained.
Fourth, this works was not free of mistakes and inconveniences. Among them it can be mentioned that it was not possible
to get the same number of subjects for the experiments as in the simulation: only a sample size 83% of the simulated one
could be recruited. Another flaw in this work is that we did not validate the simulation software, but we use LEGION as tool
to select the more promising measures to be tested in laboratory. As it was explained in this article, both problems make it
not possible a vis-à-vis comparison between simulation and experimental results, which was one of the initial objectives of
this work.
Further work on this line of research will take care of the errors mentioned above. In addition, we are already planning
more experiments in a full-scale model of a metro car in the HDL. In these experiments other pedestrian simulation
approaches will be tested and the results of the simulations will be compared with those obtained from laboratory
experiences.