I've started to add animation state information into the behavior tree. Right now when a patient waits forever they switch to an idle animation. These animation updates needs to be integrated into all behaviors.

See crowdedjs/assets#2

Work on speech bubbles

The draw callback (index.html) is not being used anymore. We could eliminate it or integrate the cade in viewer into this callback.

Make sure that we add non-determinism back into the simulation.

6 am and 6 pm are shift changes (tentatively), 12 hour shifts. The greeter nurse only works from 9 to 5. There are one of three arrival times. Morning, Afternoon (swing), and Evening.

See crowdedjs/math#1

In almost all places where we use distanceTo and length, we could use distanceToSquared and lengthSquared. Times include when we compared based on distance (x^2>y^2=>x>y) such as this updated line:

https://github.com/bricksphd/trash.recast-detour/blob/e9100c3f542e19f75c0314b82866d688d1d43157/src/behavior/look-for-arriving-patient.js#L28

Or when we compare to a constant such as this updated line:

https://github.com/bricksphd/trash.recast-detour/blob/e9100c3f542e19f75c0314b82866d688d1d43157/src/behavior/look-for-arriving-patient.js#L30

The only time we really need to use length instead of lengthSquared is when we normalize, so we need to keep it in for that.

The patient and tech circle endlessly

Describe the bug
The reported simulation fps can be inaccurate.

Expected behavior
The simulation fps is the actual fps.

Wanted to note here while I'm thinking of it, this is seemingly the case for Sev A patients and normal patients, as there are several spots where having multiple medical agents doesn't help for one reason or another. Some of those reasons include:

• When there are Sev A patients that need to be seen in the CT rooms, the tech that is NOT on the A team stops working (assuming all vitals have been taken already) and lets the A team tech take and retrieve the Sev A patients from the cat scan before they start helping again.
• There is a choke point for the Resident agents too. Resident agents rely on Techs to do their tasks before they do theirs, so when the techs are constantly moving across the whole hospital floor to take patients to the CT Rooms, it often means only one Resident is active at a time, as there is not enough for them to do until the techs accomplish their tasks. For instance, the tech must take vitals and ekg before the resident can consult with the patient.
• I believe that some of the patients/acknowledgements get piled onto one of the two medical agents of each type disproportionately sometimes, which means often throughout the simulation one of the two agents will be doing nothing when clearly they could technically be doing something, if only it were assigned to them.

Patients often require:
A CAT scan
An EKG
An x-ray
A blood draw.
I believe we have the first two implemented, but not the latter two.

Any sev 1 patients automatically should be put at the front of the queue.

Putting this here so I remember it later.

I think that we're tracking location in two places (my fault). I think we have the locations array and a location (which is the last frame). I think the location variable is redundant since the information is already in the locations array.

Delete unused and old code branches, try to narrow it down to main (my current build), and a dev branch (future branch).

Right now the frame controller has play (reverse), pause, play (forward), and three different speeds (x1, x5, x10).
I think the frame controller should also have a back one frame, forward one frame, go to beginning, go to end, and track last frame options.

We lost animation blending in the most recent version. We need to put that back in.

I wrote the code to throw exceptions incorrectly. The exceptions should read:
throw new Exception(string)
not
throw new [string]

Right now all people (including patients) are created in the Medician object. Patients should be created in the patient object. This will allow us to leverage OOP better.

We need to have patients leave two ways--the way they came and by being booked into the main hospital.

In the assets repo, agents have their arrival encoded as a tick, e.g.
https://github.com/crowdedjs/assets/blob/c0841be5261f3762533a3a8c143a398b7f963619/arrivals/arrivalHospital.js#L4

Agents are spawned in crowd-setup based on their .startMSec attribute,
https://github.com/crowdedjs/crowd-setup/blob/12257fe45d6ad6abd108d7acfdcebd7c8d27184a/index.js#L75

The conversion from ticks to milliseconds take place in the agent constructors, e.g.

https://github.com/crowdedjs/examples/blob/0fea4001d99c3e1e41659ca0714010e40cda8e37/src/index.js#L59

Commenting out this line makes the simulation run better, so clearly some optimizations need to be made!

Add a global computer to the simulation.

When a sev 1 arrives and is sent to an A bed, there is usually already a team assigned that will drop what they are doing to assist. This usually includes a doctor, a resident, a nurse, and a tech. Often phlebotomy will send someone and someone from imaging will also be on hand.

A real ER is teaming with doctors, residents, nurses, and techs. Right now we have only one of each.

The crowd speeds up after a shift change, suggesting a memory leak.

In alex-branch agents' animations were throttled by the distance the agent travelled. We lost that functionality when I merged.

Sev 1 patients often arrive at the ambulance drop off. We need to handle this situation.

We should be using the actual arrival data that represents an average day at the ER

After some changes a long while ago, our scoring functions stopped working and outputs multiple errors into the console when the simulation starts. It looks like it's not something I can just comment out on my end as it is apart of the node_modules folders.

examples/node_modules/@crowdedjs/crowd-setup/index.js

(The errors was at line 188, "can't access property "id", agentPositionsRef[j] is undefined")