Do you try to write code without knowing where you're going first? If so, try to spend some time with pen and paper and write/doodle whatever you need to in order to get it clear in your mind what you need to write. Crude diagrams of flow control, pseudo-code, even just defining the problem you're trying to solve. Break down the problem into smaller problems. Then turn some of the smaller problems into data and/or code. For example:

I need a program that simulates a traffic light sequence. I need to represent a box with three lights that can be on or off in memory. I need to move from one state to the next, so I need a way to find the next state using the current. I need to loop forever. I need each state to last a certain duration. I'm eventually going to need a thousand lights in my city simulation, starting at random states.

// Represent lights and states in memory.
typedef struct { int r, a, g, duration; } LightState;
typedef struct { int state; ... } Light;

// Represent all possible states and transitions.
LightState states = {
  {1, 0, 0, 60},
  {1, 1, 0, 5},
  {0, 0, 1, 60},
  {0, 1, 0, 5},
};
int transitions = { 1, 2, 3, 0 }; // next = [current]

// Create a light and change states forever.
Light light = {0};
while (1) {
  sleep(states[light.state].duration);
  light.state = transitions[light.state];
}

I'm sure you could imagine two flow diagrams you could draw before you got to this, e.g. the states of the light and the sleep->change->sleep->change loop, and maybe some pseudo-code for the loop etc.

Now that I've solved this part I can create my thousand lights in a loop, using rand() or similar for the starting states. Then each tick of the above loop I'd iterate over all lights, running the above code, and so on...

Try not to program "by accident" or "by coincidence" and never write too much code without running and testing it.