PAXLink - a PAX-only AI that uses feeder systems

[Drury]

I loaded one very old saved game but then PAXlink (yellow) built 1824 stations... and without any buses

[Zuu]

Do you have (and can upload) this old save game? In the old save game did you have PAXLink there or was there another AI playing initially?

Which version of PAXLink do you have?


Interesting bug though. :-p

[Drury]

The only AI's bus is made by me...

[Zuu]

Thank you. Whenever I get time to work on PAXLink, I'll take a look. Note however, that PAXLink is not very high on my list currently.

It is probably related to pre-existing infrastructure not built by PAXLink that it misinterpreted. As always, if more people experience this problem, please put a note here as that gives me a picture of how big the problem is in practice. So far my guess is that this is a very rare problem.

[Kogut]

Rather not the best strategy (1950, no newgrf except eyecandy)

[Samu]

Sem nome, 1985-03-20.png

(202.07 KiB) Downloaded 1 time

Hi. The AI just started, placed a statue, 4 bus stations (2 in each town), 54 buses (27 in each town), 1 airport and 0 planes.

The screenshot is really scary!

[Zuu]

Due to the huge size of the city (with no other passenger servicies), it has calculated that it needs a lot of buses to provide good service. If you let it continue for a while it should detect that the bus stops are overloaded and reduce the amount of buses. I don't remember now if PAXLink has the capability of expanding the feeder stations, but if it has, then it will try to expand the bus stations with more stops. I don't remember if I made it so it is allowed to destroy any buildings or not.

[Samu]

It went bankrupt without selling any of the buses.

I think building a statue in town too early will make the AI have financial problems right away.

This is one of the AIs I have most trouble finding the best map/AI definitions so it can survive. It has never succeded for me.

[Zuu]

This AI is more a test concept than a 1.0 product. The part where it lacks most is probably to weight different investments against each other. I've started some work on making it calculate a gain/cost value and how long time it will take for it to get the money required for different investments. Then the plan is to choose an action based on these two values and possible some other factors. However this is quite a large task as the AI have many different actions that needs a cost and profit estimation.

Just finding out how much money the company makes in a year or so is non-trivial. For example the AI can't query OpenTTD for the maintanence costs of stations. This is a fee of about 500 pounds a year in 1950, but depending on inflation and possible BaseCost NewGRFs it could differ over time and from game to game.


That said, currently PAXLink is quite low on my priority list of various hobby projects. It is currently more likely that some ideas form PAXLink finds its way to CluelessPlus than the other way. So far there has already been some movement of ideas/code from PAXLink to CluelessPlus, including the statistics module that keeps track of the usage of stations.


A way to make PAXLink get started is to use the cheats to give it some money. Starting in 1960+ is usually also of benefit as I've not implemented upgrading of airports.

[MAG101]

Nice AI is there Future Release Impovements in the future?

[Zuu]

At the moment I don't work on PAXLink anymore. There are some unpublished changes for it where I've started to break apart the AI into tasks which all have an estimated cost and an estimated benefit. There is a long way to go on this work which is one of the reasons why I've not got anywhere further.