Videos show path finding inherently "broken"

[Dirktator 2,305]

Some videos demonstrating that the path finding is at best disappointingly primitive.

Really surprised and dismayed.

[Darth Apple 54]

It seems like the sims are taking the shortest route possible again. Glad that was fixed for SC4, but it's really pretty sad that after ten years since SC4 the same problem comes with SC2013. From the looks of it, it seems like even a curved road will bump the travel distance just enough to cause sims to use another route.

 

Hopefully they fix this pretty soon. It doesn't seem like they really learned their lesson from Simcity Societies.

[Bobington 4]

I thought this was common knowledge

[daneoid 78]

Everything in this game is simulated by an individual agent......poorly.

[magicrobbie 77]

Thanks - clearly demonstrates that there are a lot more issues than just server problems/DRM. If pathfinding fails on such a basic level, the whole idea behind Glassbox (agent-based simulation) falls apart. Epic failure! What were they thinking and doing?

[Darth Apple 54]

EA is definitely going to need to really work hard at fixing a lot of this. Nobody ever expected anything to be perfect, but this is beyond flawed.

 

I still believe the game is redeemable, but EA is going to have to go against their track record in the past to make it happen. We'll have to see what happens.

[Ohadi195 30]

Everything in this game is simulated by an individual agent......poorly.

 

lol that's just horrible!

 

It's like the traffic always flow in the shortest way (even under high congestion).

In Cities XL the traffic flows in the fastest way - hence the congestion much more balanced and realistic. heck, CXL doesn't have the glassbox engine.

 

I can only imagine what would happen when you build an Highway. the traffic will still use the congestioned roads because it's the "shortest" way to drive. even if the highway has 8 lanes and higher speed limits. 

 

it's just sad, really sad...

[Motina 77]

I noticed this in one of their videos. Can't seem to pin point it now, but but it had them setting up a hospital, and said, and there goes the ambulance... away from the injure/sick sim.

 

Sad.

[jacksunny 741]

I've typically been very positive about this game and its features, but this is just ridiculous, and completely unacceptable. 

[pfudke 6]

This reminds me how inefficient garbage trucks are.

 

 

[croxis 150]

IIRC fromt he GDC glasbox videos the agents use a modified A* pathfinding that searches the shortest route to the closest possible "need" for that agent.

[Mr. Flibble 134]

Aren't problems like these what the $19.99 "Streets and Rails" addon that you just know will be coming will allegedly remedy?

[Dirktator 2,305]

IIRC fromt he GDC glasbox videos the agents use a modified A* pathfinding that searches the shortest route to the closest possible "need" for that agent.

 

What's disappointing is that the paths could be weighted based on traffic density so that shortest route isn't just the only criteria.

 

Further, Maxis once stated that agents would start out with the intention to travel the shortest route but would update their path finding based on new data such as alternative routes on roads that can support higher density traffic.  What's shown in the videos is clearly not the case.

[Ohadi195 30]

IIRC fromt he GDC glasbox videos the agents use a modified A* pathfinding that searches the shortest route to the closest possible "need" for that agent.

 

What's disappointing is that the paths could be weighted based on traffic density so that shortest route isn't just the only criteria.

 

Further, Maxis once stated that agents would start out with the intention to travel the shortest route but would update their path finding based on new data such as alternative routes on roads that can support higher density traffic.  What's shown in the videos is clearly not the case.

 

I can't get it. how did Monte Cristo have managed to do so on Cities XL (and must say in an incredible way) and Maxis did not? is it that complicated?

[el.lopan 54]

Aren't problems like these what the $19.99 "Streets and Rails" addon that you just know will be coming will allegedly remedy?

 

Bingo.

 

Basically, the pathfinding algorithm is like:

 

1- Find the shortest possible route.

2- Ignore all other options and variables.

[Board_Marshal 104]

Since I'm no programmer, software engineer, or developer so I'm going to speculate.  Is it not possible to write a code which says:

 

First tier: distance and rate of travel
Go from A to Z by shortest route.
If travel on road B falls below x rate then at next B*C intersection check rate of travel on road C, take road C if faster or stay on road B if slower.
Go from current point A1 (either still on road B, the same route as before or changed to alt road C) to Z by shortest route
If travel on road C falls below x rate then at next C*D intersection check rate of travel on road D, take road D if faster or stay on road C if slower.

Second tier: factoring in size of alternative roads
At the intersection of road B*C check the rate of travel on road C AND check the size of the alt road C (throughput). 
If the rate of travel on C is = B and alt road C has the same throughput of road B: stay on B
If the rate of travel on C is = B but alt road C is larger: 50/50 chance sim turns onto C or stays on B
If the rate of travel on C is = B but alt road C is smaller, stay on B.
If the rate of travel on C is > B but alt road C is smaller than road B, the increase in the rate of travel on alt road C over B must be of Y amount or stay on the larger road B.
If the rate of travel on C is > B and alt road C is larger then take alt road C
If the rate of travel on C is > B and alt road C is the same size: 50/50 chance sim turns onto C or stays on B
If the rate of travel on C is < B and alt road C is smaller then stay on road B
If the rate of travel on C is < B and alt road C is larger than road B, if the decrease in the rate of travel on alt road C < Y then take alt road C, if not then stay on B.
If the rate of travel on C is < B and alt road C is the same size then stay on B.

Third tier: factoring the distance if alt road C is taken AND travel time of that path versus the current distance AND travel time.
At the intersection of road B*C check the rate of travel on road C AND check the size of the alt road C AND then calculate if, hypothetically, the sim took alt road C the time to travel to Z as a function of the shortest route from point A1 and Z and the current rate of travel on each of the roads between A1 and Z, compare to the current path (initial shortest route from A to Z) and the current rate of travel on all the roads from A1 to Z of the current path.

 

This type of system would lead sims to take shortest route where possible.  Next, if the shortest route (in terms of distance) was being negatively impacted by speed then the sim would look to take faster route.  Next, the sim will try and stay on larger roads because higher throughput likely means faster speeds but may sometimes take smaller roads if the rate is significantly faster than a jammed up larger road.  Next, since always taking the faster road may lead to a growing distance between the current point and destination the glassbox will quick calculate if the net gain of speed and road size is worth the cost of additional distance of the alt route, if yes, then the sim will drive farther overall but still get there at least as fast or faster than before if the sim had stayed on original or previously calculated shortest route.

 

There are likely additional factors the glassbox could calculate as well and this could be simulated and each tier “weighted” to when there are multiple benefits and negatives of equal numerical value but not necessarily efficiency or desirability.  Is the glassbox not capable of this?  Wasn't this the purpose of the always-online because their servers we acting like super-computers executing massive numbers of calcuations for a massive number of individual agents constantly.  Am I just being ignorant?  I do some statistical programming with lots of "if then this then do, else" and "do loops" but not software programming or development.  Are these type of path finding systems not easily written or tested?  Am I crazy?

 

[robcards 1]

It should be a very simple code to determine the quickest route from point A to point B with traffic and street capacity factored in. The code for all of what you said is simple a series of if statements. However, I am not too familiar with the code for simulating the AI, and I'm guessing the issue is in that, like it's calculating the fastest route for every sim at once and not adjusting the route after its already calculated.

[nataniels3 61]

The agents are a metaphor for EA. They look for the shortest possible route ( no subways, small cities, server-maggedon) which isn't always the best. Don't qoute me on server-maggedon, its a hyperbole.

This reminds me how inefficient garbage trucks are.

 

 

The agents driving those garbage trucks need to be fired. Those trucks drive up the street and down the street, and there aren't any houses or buildings on the street.

[Board_Marshal 104]

It should be a very simple code to determine the quickest route from point A to point B with traffic and street capacity factored in. The code for all of what you said is simple a series of if statements. However, I am not too familiar with the code for simulating the AI, and I'm guessing the issue is in that, like it's calculating the fastest route for every sim at once and not adjusting the route after its already calculated.

But I thought that the marvel of "Glassbox" was that each agent acted independently so shouldn't each sim have its own set of calcuations which are made independent of the decisions being made by every other sim at the same time.  Since no 2 Sims can possess the same "space" at  the same time each calculation will be different because the values of the parameters will be different, even if only by decimal points.  If the Sim 5 cars ahead turns onto a road and changes the congestion value of that road then the Sim 5 cars back, when he/she reaches the same previous point then the decision to turn onto the same road will be based on new values of the same algorithm, yes?  Perhaps not taking the turn an instead going a different way?

[ifurkend 0]

I can't get it. how did Monte Cristo have managed to do so on Cities XL (and must say in an incredible way) and Maxis did not? is it that complicated?

 

In CXL the animation of the vehicle traveling is pre-calculated statistically. If you've modified the network, the vehicle will not get stuck but re-display the revised solution like changing destination or travel on another possible path.

 

However, in SC2013 where every vehicle (agent)'s path is calculated in real-time environment, it's very reminiscent of Transport Tycoon (TT) that although every vehicle's path is pre-calculated according to your assigned destinations, how each vehicle reacts in the network and what materials the vehicle is carrying are all calculated in real-time. In TT your trains can be jammed at intersection PERMANENTLY if more trains share the same network. Also your train can get lost in the network and there is one common cause to that that is very similar to the SC2013 pathfinding issue:

When the train is waiting before a signal with a branch to other destination is avaialbe, if the train has waited for too long it will attempt to detour irresponsibly by going into that branch regardless of the possibility to get lost forever. (To add insult to injury, this was not fixed even in TT's spiritual successor "Chris Sawyer's Locomotion" 10 years later.) In SC2013 this is slightly better that every vehicle can u-turn at the deadend. In fact, the vehicle in SC2013 detours is possibly due to the lane before intersection requires the vehicle to queue for too long so it chooses the other lane which leads to a different road to avoid congestion. Also it helps to relief the traffic behind to "make good use" of road/lane which still has capacity to carry traffic.

I'm not defending for Maxis' programmers. The pathfinding of SC2013 is definitely flawed. But I can see that any solution to that must not increase the data flow exponentially. In the very first version of TT the quota of vehicle quantity of each type (road vehicle/train/ship/airplane) is very limited, and this cap is shared by all participants in the same map, including the AI opponents.

[Catchin22 299]

Some things in here are notable mistakes but again, blown out of proportion. How many people IRL take the best route 100% of the time? In my town there's a road that hits a red light and the wait is fairly long in order to release traffic from a parallel road, which often doesn't actually have any traffic. You would be surprised how many people sit and wait on that road backing traffic up a half a mile instead of driving down another street crossing over into the other lane with the green light.  The coding going into path finding is not going to be any wear NEAR as simple as you think. Especially when you start factoring things like trash removal, it can all get fairly complex.

[Board_Marshal 104]

Some things in here are notable mistakes but again, blown out of proportion. How many people IRL take the best route 100% of the time? In my town there's a road that hits a red light and the wait is fairly long in order to release traffic from a parallel road, which often doesn't actually have any traffic. You would be surprised how many people sit and wait on that road backing traffic up a half a mile instead of driving down another street crossing over into the other lane with the green light.  The coding going into path finding is not going to be any wear NEAR as simple as you think. Especially when you start factoring things like trash removal, it can all get fairly complex.

The issues being discussed may or not be being portrayed in a manner that is "proportional" to some unknown benchmark of seriousness but I'd argue that nobody has ceded authority to you to decide what is "proportional" but I still respect your opinion regardless of whether or not I agree with it.

 

Since I can't talk about all people IRL I will only speak about myself.  I considered the shortest route to work (and most places) and then made a lot of assumptions based on past experience and then drove accordingly.  The route varies sometimes and the shortest route isn't always the fastest, even the day of the week can be a factor.  I could be wrong but I don't believe the Sim agents act with "experience" so it seems that any pathfinding algorithm has to be based on currently available data without "memory" of previous game behavior, e.g., the traffic was busy yesterday so the Sim will take a different route today and possibly tomorrow.

 

As I stated before I'm no programmer so I don't know if the coding for pathfinding is complex.  If you are the SimCity employee who wrote the code for pathfinding then I'll agree with you if you say the code is in fact complex.  If you're not that employee then perhaps the coding is simple, maybe it is complex but it was done poorly, maybe it is complex and the code is written well but it is not very good at achieving its desired effect.

 

I play many of the Paradox interactive games (they're not the best looking but they're known for their complexity), for example, HOI3 and there has been all sorts of stuff that wasn't very good initially which took tens of patches, several expansions, and several years to improve.  The PI devs admitted all the time that stuff didn't work out as they liked, dropped out stuff that just plain didn't work, and added new stuff, etc.  My point is that SimCity may have legitimate problems (its devs aren't perfect beings) which could be considered small, big, ridiculous, and some which are being "blown out of proportion".  At the moment there is so much going on I don't believe anyone outside of the employees at Maxis know what's actually performing poorly (beyond the server issue) or what is truly working as designed and people just need to figure it out.  Despite all of the freaking out I'm confident Maxis will share with the community the information about what's working as designed, what's not, and how they plan to stick by their product and make it better (I assume anyway).

[Mr_Maison 2,426]

There is a legitimate time to say when something is broken and at that time is beyond opinion. The #1 way to tell is when a thing cannot perform the task that was intended. If the the end result is because of user ignorance, then perhaps it's not broken but the user is not "doing it right". But if all steps to use or operate something is done correctly as directed, then truthfully either the thing is broken or the design is bad in the first place. If the design was bad in the first place then I wouldn't say it's broken. For something to be broken it would've needed to be operable at some point in it's existence. What I'm seeing here is just bad ideas that are already dead on arrival due to bad design choices. The architect is the true master builder and the success of a creation is in the dream before any brush hits the canvas when it comes to creating something intentionally. Someone just didn't think this one through while ignoring the dream of playing SimCity.

[el.lopan 54]

Some things in here are notable mistakes but again, blown out of proportion. How many people IRL take the best route 100% of the time? In my town there's a road that hits a red light and the wait is fairly long in order to release traffic from a parallel road, which often doesn't actually have any traffic. You would be surprised how many people sit and wait on that road backing traffic up a half a mile instead of driving down another street crossing over into the other lane with the green light.  The coding going into path finding is not going to be any wear NEAR as simple as you think. Especially when you start factoring things like trash removal, it can all get fairly complex.

 

If you want to compare this scenario to "IRL", then nor do real people take the worst route 100% of the time. If a sim were going from point A to point B far apart through a complex path of avenues, intersections and whatever, then sometimes it'd be acceptable if the best route wasn't chosen, just like IRL. On the other hand, if you got 2 roads ahead in parallel that clearly lead to the same way, the first one is completely free and the second one got a heavy traffic jam and you choose to go through the jammed one, one could only call you a masochist.

[Wiimeiser 434]

Then according to Maxis this "issue" is intentional?

[Board_Marshal 104]

Then according to Maxis this "issue" is intentional?

Not sure, I can't imagine it would be the case.  I would assume the people at Maxis have played hundreds of hours (if not thousands) and you'd assume they would have noticed this before.  My own opinion is that a big reason why people are losing their minds is because this all seems so preposterous that the only way for the human mind to process it is to flip your shite. 

 

Because the devs appear to have sunk so much into the travel network being the linchpin by which all of the "agent" activity occurred you'd think that pathfinding along the travel networks would be the absolute number one priority.  Perhaps it was number 1 but the Glassbox and SimCity are so freaking sophisticated that even the devs who created it couldn't control the beast they created, maybe.

[MandelSoft 2,769]

And I thought the Sims in SimCity 4 were stupid, but this is just another disappointment about the game. I'd rather have a good statistical based simulator rather than a dumb agent-based simulator.

[RepublicMaster 428]

And I thought the Sims in SimCity 4 were stupid, but this is just another disappointment about the game. I'd rather have a good statistical based simulator rather than a dumb agent-based simulator.

I agree completely. This is really disappointing, you think they would have learned.

[ifurkend 0]

And I thought the Sims in SimCity 4 were stupid, but this is just another disappointment about the game. I'd rather have a good statistical based simulator rather than a dumb agent-based simulator.

If the agent simulation was done correctly it would be a benchmark for the next-gen simulation game which handles massive interactive individual data in real-time, much better than the traditional statistics-based simulation. Seems the plan was bit too ambitious or it was EA's order to reduce the complexity of agent algorithm so older computer could run the game/engine smoothly also.

[jumbz55 7]

Aren't problems like these what the $19.99 "Streets and Rails" addon that you just know will be coming will allegedly remedy?

Hahaha! Such a good call