In this paper, we explore the applicability of reservation concept
for congestion mitigation, particularly focused on the downtown
of a city. Although it is called downtown space reservation
system, its application is not rigorously restricted to a downtown
area. Streets of a central business district resembling an area could
use a similar approach. The proposed DSRS consists of two modules,
an offline optimization module and an online decision making
module. In the offline module, an optimization problem is solved
based on historical travel information. Multiple objectives are
incorporated into the optimization problem. The optimization
solutions show that under known demand patterns, the system is
able to pick the ‘‘best” trips and achieve the ‘‘best” system performance
in terms of people throughput and revenue generation.
Based on the optimal results from the offline module, an ‘‘intelligent”
system is developed using neural network technology. Not
only is the system able to learn and generalize from the optimal
solutions, but the system is also able to update itself, as the new
information arrives. With an extensive training, the ‘‘intelligent”
system can facilitate decision makers to make real time decisions
on whether to accept/reject a reservation request. The intelligent
system may not guarantee the optimal results compared with
the results from the offline module. However, for a known traffic
pattern, it does show consistency. In the numerical example, one
can catch a glimpse of the way that the system in principle works
and to what extent the neural network resembles optimality in the
offline module. Leaving aside the implementation aspects, the system
can be promising.