• To build, run ./build.sh
  • Compiles using C++11 (which has range-based for loops and the "auto" type)
  • If there is an error running this script, run chmod +x ./build/build.sh
• To run, run ./build/simulation t s, where t is the number of trucks and s is the number of stations
• Statistics should be reported at the end

• To test, run ./test.sh
  • Compiles with print statements turned off so unity testing is the only output
  • Compiles using C++11 (which has range-based for loops and the "auto" type)
  • If there is an error running this script, run chmod +x ./build/test.sh
• To run, run ./build/test_sim
• Statistics should be reported at the end

With 1000 trucks and 2 mining stations, here is the output:

ColinSkkpro1223:VastTest magi-nerv$ ./build/simulation 1000 2

----- Initialization -----
Requested 1000 trucks and added 1000 trucks
Requested 2 stations and added 2 stations

----- Simulation -----
Started 4320 minute simulation
Sim Stopped at 4320 min

----- Statistics -----
Most unloads: 3
Longest wait time: 2290 min
Longest mining time: 300 min
Average time spent waiting: 85.79%
Average time spent mining: 11.04%
Total mining minutes: 477092 min
Total waiting minutes: 3706039 min
Max queue length: 458 trucks

• Classes:
  • SimObject:
    • Base class for all objects in the simulation
  • Truck:
    • Has a state machine
      • States: MINING, DRIVING_FROM_MINE, DRIVING_TO_MINE, UNLOADING, and WAITING_AT_STATION
    • Keeps track of number of its own statistics
  • Station:
    • Keeps queue of Truck pointers that acts as its line of trucks waiting to unload
    • When a truck reaches the front of the queue, the station notifies the truck that it should begin unloading
  • Simulation
    • Keeps time and calls the simulate_timestep() functions
    • Has the "find station with least trucks" function for a truck to call

I ended up modeling this simulation after a CS polymorphism/inheritance project from one of my classes in which a central "Simulation" class calls the "simulate" functions of all the objects in the simulation. I also tried to avoid dynamic programming, since the only objects I needed to allocate were the trucks and stations at the beginning of the simulation.

I realized that a lot of the rest of the implementation (in addition to finishing tests) would be logging and reporting all the statistics, which falls more into data science or analysis (like with GNC sims). It would take a lot more time to fully implement all my ideas (something I may do after I submit this because this is a really fun project). I had many more ideas for statistics that I included below but have not implemented all of them yet due to time constraints.

In general, I had more time, I would include more stats that track distributions rather than single numbers (i.e. maximum wait time).

I would include several more vectors that log the state of individual trucks/stations over the course of the simulation. I already included total_action_minutes_by_state for trucks and current_truck_unloading_by_min for stations, and theoretically, these could be used to derive all other stats. Here are some other stats that are useful.

Possible Structures for Truck

• Vector of mining times

  • e.g. If a truck mined twice first for 60 minutes, then for 120 minutes, it would create --> map[MINING] = {60,120};
  • Used to derive individual mining distributions
  • Can be used to determine
    • Truck with most mining sessions

• Vector of Station* at each unload

  • Which truck stopped at the most or fewest different stations.

Possible Structures for Station

• Vector of Truck* unloading at each minute
  • Can be used to determine distribution of which different trucks stopped at that station

Possible Structutes for Simulation

• Number of trucks in a given state per minute
  • e.g.
    • std::vector trucks_mining_by_min
    • std::vector trucks_waiting_at_stations_by_min
    • std::vector current_truck_unloading_by_min
    • std::vector trucks_going_to_mines_by_min
    • std::vector trucks_coming_from_mines_by_min
  • Can graph distribution of trucks at each state over time

I would also add more stats that are helpful in assessing efficiency of the unloading stations. I already included percent_of_time_mining for each truck, which helps in maximizing the efficiency of an individual truck. One unimplemented stat might be min_waiting_time, which could be used to determine how many trucks had to wait

I definitely think that I have a good intution for how to create a simulation such as this, with the central Simulation class running the whole program. But there are things I would improve for next time:

• Better testing (including a more test-as-I-go approach)
  • Includes modules that are better designed
  • With hindsight, I may change the simulation to add trucks in different locations at the start of the simulation (i.e. waiting in line) for better testing.
• Data validation for inputs
  • I didn't really have enough time
  • This also goes hand-in-hand with better testing, as this makes modules have more defined unit tests for each function
• Better documentation as I write my code
  • I try to name functions and variables so that they are easy to understand, but this also includes commenting
  • Since I was changing functions often, I opted out of documenting as I went, but ran out of time to finish it completely.

I believe my solution is fairly intricate and scaleable for more types of objects than just trucks and stations. I'm proud of the implentation of the unity testing, and that is something I would like to continue improving for more robust code, especially this summer working on Haven-1.

Thanks for this great opportunity!