The decision to look for a lane change depends on traffic conditions, driver's destination and behavior characteristics (e.g. from conservative to aggressive). MITSIM classifies lane changes into two types: mandatory and discretionary. Mandatory lane changing occurs when drivers have to change lanes in order to:
If a downstream node has multiple outgoing links, vehicles approaching that node are tagged as having a mandatory state according to a certain probability. Before a vehicle is tagged with mandatory state, both the left and right lanes can be the target lane of a discretionary lane change. After being tagged to mandatory state, the vehicle can seek two types of lane changes: (i) if the current lane is not connected to the next link on its path, it will make a mandatory lane change or a series of mandatory lane changes so as to move onto a lane leading to the next link; and (ii) if the current lane is connected to its next link on the path, then it can only make discretionary lane changes onto lanes that also connect to its path. The types of lane changes and their relationship with mandatory state are depicted in Figure 3.6. In the following, we describe how the probability used to tag vehicles to mandatory state is determined.
Figure 3.6: Mandatory state and lane change
types
Assume that, at a particular position and congestion level, the probability a vehicle should be tagged to mandatory state can be written as:
where:
;
is a variable defined
as follows: where:
As depicted in Figure 3.7(a), the probability given by Eq (3.15) increases to 1 as the vehicle approaches the critical position. When the number of necessary lane changes and traffic congestion increases, this probability also goes up as shown by the different curves in Figure 3.7(a).
Figure 3.7: Probability of tagging vehicles to
mandatory state
The simulator draws a random number in each iteration to decide
whether an untagged vehicle should be tagged to mandatory state. Let
be the probability to tag a vehicle at time
interval i. Then its relationship with 
- the
probability that the vehicle should have been tagged upto
ith time interval - can be represented recursively as the
following:
Since 
, 
, 
, 
and are known at the
ith time interval, can be written as:
An example of the results obtained from Eq (3.18) is illustrated in Figure 3.7(b). The curves in this
figure will shift down if the simulation step size decreases, and vice
versa. In other words, 
obtained from Eq (3.18) depends on the
sampling rate. This is to guarantee that the probability a
vehicle has been tagged at a given condition is independent from the
simulation step size.
When a vehicle has been tagged to mandatory state, it keeps that state until it has performed the desired lane change or moved into the downstream link.
For discretionary lane changing, the decision to stay or change is based on traffic conditions of both the current lane and adjacent lanes. If a vehicle has a speed lower than the driver's desired speed due to a slow vehicle in front or the maximum speed of that lane, it checks the neighboring lanes for opportunities to increase its speed. Several parameters, including an impatience factor and a speed indifference factor, are used to determine whether the current speed is low enough and the speeds in adjacent lanes are sufficiently high to consider making a lane change.
In order to decide on the lane to move into, the vehicle first
determines the set of admissible lanes. A lane is defined as
admissible based on several criteria including lane changing
regulation, lane use privilege, lane connectivity, incident, lane use
signs, message signs, prevailing traffic conditions, driver's desired
speed, lane's maximum speed 
, and whether the
vehicle is in mandatory state.
