Evaluation

Evalutation Metric: Average Travel Time

For those vehicles which complete the trip from the starting road to the ending road, we define the timestamp it reaches the destination (enter the last road segment) as $t_e$. Then the travel time is defined as the time difference between its starting time $t_s$ and ending time $t_e$ and it is calculated by:

$$ \tau = t_e - t_s $$

As for the rest of the vehicles which have not reached the destination when the system time is over at timestamp $T_{max}$, their travel time is calculated as:

$$ \tau = T_{max} - t_s $$

Then for one test scenario, the metric Travel Time is calculated as the average travel time of all vehicles where $n$ is the number of vehicles.:

$$ T = \sum_i\tau_i/n $$

Obviously, the objective of this problem is to seek a better signal timing plan to minimize the average travel time $T$.

Overall Score

For each test scenario $i$, we have a baseline travel time $b_i$, then the score of the i-th scenario is computed as the relative improvement over the baseline results: $$ score_i = \frac{b_i-T_i}{b_i} $$

The baseline result is based on the result from SOTL algorithm, which is provided in the sample code.

For $n$ different scenarios, the overall score is: $$ overall\ score = \sum_i score_i/n $$

References

  1. Getting Started with CityFlow Simulator
  2. Traffic Signal Controller Sample Code
  3. Signal Timing Manual - Second Edition, TRB’s National Cooperative Highway Research Program (NCHRP) Report 812