In other languages: 简体中文

Tutorial:Transport use cases: Difference between revisions

From Official Factorio Wiki
Jump to navigation Jump to search
m (formatting)
mNo edit summary
(15 intermediate revisions by 8 users not shown)
Line 1: Line 1:
{{sublinks}}
{{Languages}}
This page compares the three main [[transport]] methods: [[Belt transport system]] (conveyor belts), [[railway]] (trains) and [[Logistic robot|logistic robots]]. Note, that there are two more useful methods to transport items: The characters [[Inventory]] and the [[car]]. (Especially the car has an extremely high capacity and can be used instead of trains - of course you need to drive everything yourself, see the chapter about the car).
{{DISPLAYTITLE:What Transport for which case?|noerror}}
This page compares the three main [[transport]] methods: [[Belt transport system]] (conveyor belts), [[railway]] (trains) and [[Logistic robot|logistic robots]]. There are two more useful methods to transport items: the character's inventory and the [[car]]. (Especially the car has an extremely high capacity and can be used instead of trains. Of course you need to drive everything yourself; see the chapter about the car).


The best choice depends on the context in which they are used. For example on the amount of resources you need to transport, or how far away they are.
The best choice depends on the context in which they are used. For example on the amount of resources you need to transport, or how far away they are.


== Comparison of Belts, Trains and Logistic robots ==
== Comparison of belts, trains and logistic robots ==


{| class="wikitable"
{| class="wikitable"
Line 20: Line 21:
| Space required
| Space required
| Small for simple products, large for complicated product chains
| Small for simple products, large for complicated product chains
| Large due to train stations and bends. There is a difference in using [https://forums.factorio.com/viewtopic.php?f=5&t=35806 Rollon-Rolloff-Stations (lot of space) or Terminus (much smaller)].
| Large due to train stations and bends. There is a difference in using [https://forums.factorio.com/viewtopic.php?f=5&t=35806 roll-on-roll-off stations (lot of space) or termini (much smaller)].
| Small for complex products.
| Small for complex products.
|-
|-
| Optimization  
| Optimization  
| Highly optimizable. Source of never ending fun.
| Highly optimizable. Source of never ending fun.
| Seldom really needed - better rebuild or built more tracks. For larger networks you can get some trouble with deadlocks/gridlocks, if not optimized, using chain signals. Optimize train stations: connect either with belts (complex) or with logistic bots (recommended).  
| Seldom really needed better rebuild or build more tracks. For larger networks you can get some trouble with deadlocks/gridlocks, if not optimized, using chain signals. Optimize train stations: connect either with belts (complex) or with logistic bots (recommended).  
| Optimize placing of roboports (for charging) and used number of ports vs. number of bots in the air.
| Optimize placing of roboports (for charging) and used number of ports vs. number of bots in the air.
|-
|-
| Parallelization
| Parallelization
| High. Two parallel belts have the doubled throughput as one. Cost is also only double. On the other hand: Afford to build that doubles also.
| High. Two parallel belts have the doubled throughput as one. Cost is also only double. On the other hand: Effort to build that doubles.
| Extreme. Two parallel tracks can transport 5-10 times more trains than one, cause if a track is blocked the trains can choose the other free track and so they don't need to stop. The afford to build two tracks in parallel is lower than doubled.
| Extreme. Two parallel tracks can transport 5–10 times more trains than one, because if a track is blocked the trains can choose the other free track and so they don't need to stop. The effort to build two tracks in parallel is lower than double.
| Not so good. Doubling the amount of bots from one to two surely doubles the throughput, but doubling robots from 500 to 1000 can result in complete chaos.  
| Not so good. Doubling the amount of bots from one to two surely doubles the throughput, but doubling robots from 500 to 1000 can result in complete chaos.  
|-
|-
| Initial costs
| Initial costs
| Very small as long as conveyor belts are short. Still good with parallel belts. Larger when making express transport belts. Expensive with express; which should be used for special cases only.
| Very small as long as conveyor belts are short. Still good with parallel belts. Larger when making fast transport belts. Expensive with express; which should be used for special cases only.
| Considerable amount of material is needed for initial locomotive. Small costs (when compared to the increase of transport) for making rails.
| Considerable amount of material is needed for initial locomotive. Small costs (when compared to the increase of transport) for making rails.
| Roboports and especially bots require a lot of resources to make. Logistic bot upgrades are very expensive, but needed and must be calculated into the total costs. Maintaining costs a lot of time.
| Roboports and especially bots require a lot of resources to make. Logistic bot upgrades are very expensive, but needed and must be calculated into the total costs. Maintenance costs a lot of time.
|-
|-
| Use of energy
| Use of energy
Line 61: Line 62:
| High throughput, bulk-goods (ores, some intermediates), small to medium distance. Examples include raw materials and simple products.
| High throughput, bulk-goods (ores, some intermediates), small to medium distance. Examples include raw materials and simple products.
| High throughput, long distance. Examples include transporting ores or plates from resource fields to main factory area.
| High throughput, long distance. Examples include transporting ores or plates from resource fields to main factory area.
| Extremely high throughput over very short distances (< 50 tiles), low to medium throughput over medium distance (50-500), catastrophic over long distance. Best used in main factory area for complex products like modules and advanced circuits. Also best used for products needed in smaller quantities like placeable structures and ammunition. Unbeatable for train station (loading/unloading chests).
| Extremely high throughput over very short distances (< 50 tiles), low to medium throughput over medium distance (50–500), catastrophic over long distance. Best used in main factory area for complex products like modules and advanced circuits. Also best used for products needed in smaller quantities like placeable structures and ammunition. Unbeatable for train station (loading/unloading chests).
|-
|-
| Most annoying problems
| Most annoying problems
| Left/right lane problems (part of the game), splitting off the right amount of items.
| Left/right lane problems (part of the game), splitting off the right amount of items.
| Deadlock situations (in complicated crossings), complex train-station and signal setup.
| Deadlock situations (in complicated crossings), complex train-station and signal setup.
| Blocking of roboports by waiting for charging, "stupid" behavior that far away bots are used, instead of near and vice versa, Robots that take a "shortage" and cross a native village.
| Blocking of roboports by waiting for charging, "stupid" behavior like far away bots are used, instead of near and vice versa, Robots that take a "shortcut" through [[Enemies|Biter]]-infested territory.
|}
|}


Line 74: Line 75:


* Practical up to distances of 500 tiles (to compare: a radar station watches up to 200 tiles, with a 100-tile "radius").
* Practical up to distances of 500 tiles (to compare: a radar station watches up to 200 tiles, with a 100-tile "radius").
* Useful for connecting small resource-fields to the factory area, as the capacity of a belt is limited to about 700 items per minute in the best case, when using basic belts. With fast or express belt, this is higher (up to 1750) but the price is gigantic. Using multiple parallel belts is costly, but with basic belts still affordable over long distances.
* Useful for connecting small resource-fields to the factory area, as the throughput of a belt is limited to 800 items per minute on a fully-compressed basic belt. With fast or express belt, this is higher (up to 2400) but the price is gigantic. Using multiple parallel belts is costly, but with basic belts still affordable over long distances.
* Best for working within an automated facility for '''bulk goods''', eg. transporting ore, plates, circuit boards, etc over short or medium distances.  
* Best for working within an automated facility for '''bulk goods''', e.g. transporting ore, plates, circuit boards, etc. over short or medium distances.  
* Store, over long distances, a not to be underestimated number of items (about 12 per tile, with two lanes, so you can calculate over 500 tiles with 6000 items!). This storage is out of reach and could be counted as wasted, until it is transported.
* Store, over long distances, a not to be underestimated number of items (about 12 per tile, with two lanes, so you can calculate over 500 tiles with 6000 items!). This storage is out of reach and could be counted as wasted, until it is transported.


Note: the length of 500 tiles is just a "rule of thumb", and comes from these facts:
Note: the length of 500 tiles is just a "rule of thumb", and comes from these facts:
* It takes nearly 5 minutes (4.44) for an item to run from one end to the other on 500 tiles of [[basic belt]]s.
* It takes nearly 5 minutes (4.44) for an item to run from one end to the other on 500 tiles of [[transport belt]]s.
* A belt with 500 tiles stores about 3571 items (about a chest of resources).
* A belt with 500 tiles stores about 3571 items (about a chest of resources).
* The fact, that about after this length it makes sense to use train.
* The fact that after about this length it makes sense to use a train.


The advantages of belts are:
The advantages of belts are:
* Belts have a constant throughput rate.
* Belts have a constant throughput rate.
* Calculations are easily possible (such as "Do I need more belts to transport items from the mines?")
* Calculations are easily possible (such as "Do I need more belts to transport items from the mines?")
* Basic belts are cheap compared to tracks and and are easier to build
* Basic belts are cheap compared with tracks and and are easier to build
* You can easily see how much items are on the belt, which makes it easier to assess if you need more throughput.
* You can easily see how many items are on the belt, which makes it easier to assess if you need more throughput.
* You can use the belt as an open storage: you can store about 72 items in a length of 10 belts, when using both lanes (but for long belts this is a big disadvantage!)
* You can use the belt as an open storage: you can store about 72 items in a length of 10 belts, when using both lanes (but for long belts this is a big disadvantage!)


Line 95: Line 96:
* Items on the belts are "lost items" until they are transported.
* Items on the belts are "lost items" until they are transported.


Belts have the best short-distance throughput for continuous, direct-to-factory transportation. '''This is especially true, if you use basic belts in parallel, cause this combination has an incredible high throughput at low costs.
Belts have the best short-distance throughput for continuous, direct-to-factory transportation. '''This is especially true, if you use basic belts in parallel, because this combination has an incredibly high throughput at low costs.


=== Trains ===
=== Trains ===


* Trains should be used to get items from big and distant resource fields. Nothing can compare with it, especially, when thinking about speed. The planning is difficult and the building itself is currently tedious. (tip: use the train to build the rails, this is much faster: sit in the train and build the next rail before you, hold mouse, start driving slowly. You can use it also to build the power poles like so).
* Trains should be used to transport items from big and distant resource fields. No other means of long distance transport is as fast as trains. A single train line can easily be built; however, planning an extensive railway setup can be difficult and time consuming. One clever way of placing rails is entering a locomotive and placing the rails in front of it using the [[rail planner]].  
* Trains are for linking separate factory assemblies together over long distances. This additionally lets you 'take a ride' out to outposts that require attention.
* Trains can be used to connect separate factories together over long distances. This additionally lets you 'take a ride' out to outposts that require attention.


By smelting ore locally and ship the plates via train, you can double your wagon capacity and route directly to a storage yard or factory. See also [[Transport/Compress_by_Pre-production]].
By smelting ore locally and shipping the plates via train, you can double your wagon capacity and route directly to a storage yard or factory.
 
An interesting thread: [http://www.factorioforums.com/forum/viewtopic.php?f=5&t=6799&p=53609#53609 Long distance transport].


=== Logistic robots ===
=== Logistic robots ===


* Logistic robots should be used in a limited area with dense building placement In most cases this is the main factory area. With some updates the logistic bots can handle enormous amounts of items.  
* Logistic robots should be used in a limited area with dense building placement. In most cases this is the main factory area. With some updates the logistic bots can handle enormous amounts of items.  
* Bots are terrible at long distances and bulk goods: It's not a good idea use bots for ore-transport or from/to furnaces. It is possible to use them for this task, but don't hold your breath for this to work perfectly; there are several issues with the AI that cause them to work inefficiently more often than not. Some players take that problem as challenge.
* Bots are terrible at long distances and bulk goods: It's not a good idea use bots for ore-transport or from/to furnaces. It is possible to use them for this task, but don't hold your breath for this to work perfectly; there are several issues with the AI that cause them to work inefficiently more often than not. Some players take that problem as a challenge.
* They do a perfect job on bulk-goods, if they can charge between their jobs, which is the case on train-stations: Train comes in, train is unloaded, bots transports to the target chests, bots charge. Next rain comes in.
* They do a perfect job on bulk-goods, if they can charge between their jobs, which is the case on train-stations: Train comes in, train is unloaded, bots transports to the target chests, bots charge. Next train comes in.
* Bots do a perfect job if you use them for random transport of expensive products; transport intermediate products to where they're needed within the local network, especially if it's completely impractical to route a belt for the job.
* Bots do a perfect job if you use them for random transport of expensive products; transport intermediate products to where they're needed within the local network, especially if it's completely impractical to route a belt for the job.


Line 119: Line 118:
== Seeing from throughput side ==
== Seeing from throughput side ==


A [[Basic transport belt]] cannot transport much more than 700 item/minute and an [[Express transport belt]] (the fastest) not more than 1750 (see above). Now compare that with a locomotive: a [[Cargo wagon]] can deliver 2000 items (ores) '''per wagon'''! And if you smelt the resources at your outpost and transport intermediate plates (copper, iron) you can transport 4000 per wagon. So when speaking from throughput, you need to swap to train or make multiple parallel belts.
A (basic, yellow) [[transport belt]] cannot transport more than 800 items per minute and an [[express transport belt]] (the fastest) not more than 2400 (see above). Now compare that with a locomotive: a [[cargo wagon]] can deliver 2000 items (ores) '''per wagon'''! And if you smelt the resources at your outpost and transport intermediate plates (copper, iron) you can transport 4000 per wagon. So when speaking in terms of throughput, you need to swap to trains or make multiple parallel belts.


This picture might change when you developed [[Logistic robot]]s. But that is not true! They can't be beaten for relatively small areas, especially with their researched bonuses, and they have some kind of intelligence; they try to deliver all requests as equally as possible.
One might think this picture would change when you develop [[logistic robot]]s. But that is not true! They can't be beaten for relatively small areas, especially with their researched bonuses, and they have some kind of intelligence; they try to deliver all requests as equally as possible. But bots are terrible at long distances and bulk goods. Here are some calculations, which explain the problem (all calculations are assumed with full upgrade for the logistic bots!):
 
But bots are terrible at long distances and bulk goods. Here are some calculations, which explain the problem (all calculations are assumed with full upgrade for the logistic bots!):


Needed robots to transport 3000 items per minute over a distance of 50 tiles you need:  
Needed robots to transport 3000 items per minute over a distance of 50 tiles you need:  
  Robot speed:                    3 tiles/sec (basic speed) * 1.4 (logistic speed bonus) = 4.2 tiles/sec
  Robot speed:                    3 tiles/sec (basic speed) × 1.4 (logistic speed bonus) = 4.2 tiles/sec
  Needed time for one transport:  100 tiles / 4.2 tiles/sec = 24 secs
  Needed time for one transport:  100 tiles ÷ 4.2 tiles/sec = 24 secs
  Need robots per minute:          24 secs / 60 secs = you need 0.4 bots to transport one item per minute over 50 tiles.
  Need robots per minute:          24 secs ÷ 60 secs = you need 0.4 bots to transport one item per minute over 50 tiles.
  Because of Logistic robot cargo: 0.4 bots / 5 items per transport = 0.08 bots to transport one item per minute
  Because of Logistic robot cargo: 0.4 bots ÷ 5 items per transport = 0.08 bots to transport one item per minute
  Total:                          3000 items * 0.08 bots/per item = 240 bots
  Total:                          3000 items × 0.08 bots/per item = 240 bots
240 is a high, but still very realistic number for a logistic network with 4-5 roboports.
240 is a high, but still very realistic number for a logistic network with 4-5 roboports.


Needed robots to transport 3000 items per minute over a distance of 500 tiles you need:  
Needed robots to transport 3000 items per minute over a distance of 500 tiles you need:  
  Needed time for one transport:  1000 tiles / 4.2 tiles/sec = 240 secs
  Needed time for one transport:  1000 tiles ÷ 4.2 tiles/sec = 240 secs
  Need robots per minute:          240 secs / 60 secs = you need 4 bots to transport one item per minute over 500 tiles.
  Need robots per minute:          240 secs ÷ 60 secs = you need 4 bots to transport one item per minute over 500 tiles.
  Because of Logistic robot cargo: 4 bots / 5 items per transport = 0.8 bots to transport one item per minute
  Because of logistic robot cargo: 4 bots ÷ 5 items per transport = 0.8 bots to transport one item per minute
  Total:                          3000 items * 0.8 bots/per item = 2400 bots
  Total:                          3000 items × 0.8 bots/per item = 2400 bots
 
2400 bots need a quite big amount of resources to produce, not accounted the needed research (for bots, speed/stacksize bonuses), number of roboports and the needed energy.


2400 bots need a quite big amount of resources to produce, not accounting for the needed research (for bots, speed/stack size bonuses), number of roboports and the needed energy.


That leaves conveyors and trains.
That leaves belts and trains.


Trains are faster than belts and offer better logistic control as you go between logistic networks.
Trains are faster than belts and offer better logistic control as you go between logistic networks.


Belts simply take forever, and you may wind up with a lot of asset tied up with simply being 'in transit'. You should see items sitting on a belt as "non-useable storage". Backed up belts can be a good thing for buffer space, but long belts are just inefficient, or better: You loose the control over the other end, due to the long gap; the resources on the other side could be long depleted, but you see it only five minutes later.
Belts simply take forever, and you may wind up with a lot of asset tied up with simply being 'in transit'. You should see items sitting on a belt as "non-useable storage". Backed up belts can be a good thing for buffer space, but long belts are just inefficient, or better: You lose the control over the other end, due to the long gap; the resources on the other side could be long depleted, but you see it only five minutes later.


=== Or in other words ===
The effort to squeeze out the last 20% of belt performance increases depending on distance and total amount and at some point the train will win. But in your central factory area the bots will never be beaten. On the other hand: in small areas, and for simple throughput of one type of item (iron ore to iron-plates for example) we can place some belts in parallel, which is cheap and efficient (uses no energy) and when using parallel belts can achieve enormous throughput. For later game, the [[fast transport belt]] is a good compromise between cost and efficiency, and the [[Express transport belt]]s should be use only for short sections.
The afford, to squeeze out the last 20% of belt-performance rises depending on distance and total amount and at some point the train will win. But in your central factory area the bots will never be beaten. On the other hand: in small areas, and for simple throughput of one type of item (furnace iron ore to iron-plates for example) we can place some belts in parallel, which is cheap and efficient (uses no energy) and when using parallel belts we reach also enormous throughput. For later game, the [[Fast transport belt]] is a good compromise between cost and efficiency, and the [[Express transport belt]]s should be use only for short sections.


=== Or in more other words ===
The amount of robots you need is directly dependent on the density you build. The central part of your factory should be dense, because not only throughput is important here, but also speed: How fast can an item be produced in total (including the transportation time).
The amount of robots you need is directly dependent on the density you built. The central part of your factory should be dense, cause not only throughput is important here, but also speed: How fast can an item be produced in total (including the transportation time).


== See also ==
== See also ==

Revision as of 16:07, 29 April 2018


This page compares the three main transport methods: Belt transport system (conveyor belts), railway (trains) and logistic robots. There are two more useful methods to transport items: the character's inventory and the car. (Especially the car has an extremely high capacity and can be used instead of trains. Of course you need to drive everything yourself; see the chapter about the car).

The best choice depends on the context in which they are used. For example on the amount of resources you need to transport, or how far away they are.

Comparison of belts, trains and logistic robots

Belts Trains Logistic Robots
Throughput Constant, low for one belt, precisely calculable, limited Extremely high, fast, not easily calculable but constant, nearly unlimited if enough space, High in small areas, not calculable/chaotic, unlimited if enough bots, cannot be optimized, terrible over long distances and bulk goods
Space required Small for simple products, large for complicated product chains Large due to train stations and bends. There is a difference in using roll-on-roll-off stations (lot of space) or termini (much smaller). Small for complex products.
Optimization Highly optimizable. Source of never ending fun. Seldom really needed – better rebuild or build more tracks. For larger networks you can get some trouble with deadlocks/gridlocks, if not optimized, using chain signals. Optimize train stations: connect either with belts (complex) or with logistic bots (recommended). Optimize placing of roboports (for charging) and used number of ports vs. number of bots in the air.
Parallelization High. Two parallel belts have the doubled throughput as one. Cost is also only double. On the other hand: Effort to build that doubles. Extreme. Two parallel tracks can transport 5–10 times more trains than one, because if a track is blocked the trains can choose the other free track and so they don't need to stop. The effort to build two tracks in parallel is lower than double. Not so good. Doubling the amount of bots from one to two surely doubles the throughput, but doubling robots from 500 to 1000 can result in complete chaos.
Initial costs Very small as long as conveyor belts are short. Still good with parallel belts. Larger when making fast transport belts. Expensive with express; which should be used for special cases only. Considerable amount of material is needed for initial locomotive. Small costs (when compared to the increase of transport) for making rails. Roboports and especially bots require a lot of resources to make. Logistic bot upgrades are very expensive, but needed and must be calculated into the total costs. Maintenance costs a lot of time.
Use of energy Gratis. Free. Always an object of discussion. ;) Low. The usage of fuel is currently very affordable. High. The roboports need a lot of energy and the bots take even more energy, as more are in the air. This can be a big problem, but is normally not an issue.
Items in transit vs. duration of transport Bad. 100 tiles of belts stores about 700 items if fully loaded, but takes with basic belts 56 seconds for transport. With express belt, this is only 3 times faster. Very good. Lot of items transported in very short time. Chaotic. When jammed it can happen, that the most important items are unreachable in the air.
Maintenance costs None Some fuel needed Considerable amount of electricity needed
Pollution None Some None, indirectly caused by use of electricity.
Best used for High throughput, bulk-goods (ores, some intermediates), small to medium distance. Examples include raw materials and simple products. High throughput, long distance. Examples include transporting ores or plates from resource fields to main factory area. Extremely high throughput over very short distances (< 50 tiles), low to medium throughput over medium distance (50–500), catastrophic over long distance. Best used in main factory area for complex products like modules and advanced circuits. Also best used for products needed in smaller quantities like placeable structures and ammunition. Unbeatable for train station (loading/unloading chests).
Most annoying problems Left/right lane problems (part of the game), splitting off the right amount of items. Deadlock situations (in complicated crossings), complex train-station and signal setup. Blocking of roboports by waiting for charging, "stupid" behavior like far away bots are used, instead of near and vice versa, Robots that take a "shortcut" through Biter-infested territory.

Experts

Belts

  • Practical up to distances of 500 tiles (to compare: a radar station watches up to 200 tiles, with a 100-tile "radius").
  • Useful for connecting small resource-fields to the factory area, as the throughput of a belt is limited to 800 items per minute on a fully-compressed basic belt. With fast or express belt, this is higher (up to 2400) but the price is gigantic. Using multiple parallel belts is costly, but with basic belts still affordable over long distances.
  • Best for working within an automated facility for bulk goods, e.g. transporting ore, plates, circuit boards, etc. over short or medium distances.
  • Store, over long distances, a not to be underestimated number of items (about 12 per tile, with two lanes, so you can calculate over 500 tiles with 6000 items!). This storage is out of reach and could be counted as wasted, until it is transported.

Note: the length of 500 tiles is just a "rule of thumb", and comes from these facts:

  • It takes nearly 5 minutes (4.44) for an item to run from one end to the other on 500 tiles of transport belts.
  • A belt with 500 tiles stores about 3571 items (about a chest of resources).
  • The fact that after about this length it makes sense to use a train.

The advantages of belts are:

  • Belts have a constant throughput rate.
  • Calculations are easily possible (such as "Do I need more belts to transport items from the mines?")
  • Basic belts are cheap compared with tracks and and are easier to build
  • You can easily see how many items are on the belt, which makes it easier to assess if you need more throughput.
  • You can use the belt as an open storage: you can store about 72 items in a length of 10 belts, when using both lanes (but for long belts this is a big disadvantage!)

Disadvantages:

  • Limited transport capacity, due to the limited throughput.
  • Slow. This is a problem with very long belts because it takes a while to fill the belt. The items on the belt cannot be used. A belt with length of 100 holds 714 items for 56 seconds. With trains this is much lower (10 seconds) and as long as they are not loaded into a wagon, they still can be used.
  • Items on the belts are "lost items" until they are transported.

Belts have the best short-distance throughput for continuous, direct-to-factory transportation. This is especially true, if you use basic belts in parallel, because this combination has an incredibly high throughput at low costs.

Trains

  • Trains should be used to transport items from big and distant resource fields. No other means of long distance transport is as fast as trains. A single train line can easily be built; however, planning an extensive railway setup can be difficult and time consuming. One clever way of placing rails is entering a locomotive and placing the rails in front of it using the rail planner.
  • Trains can be used to connect separate factories together over long distances. This additionally lets you 'take a ride' out to outposts that require attention.

By smelting ore locally and shipping the plates via train, you can double your wagon capacity and route directly to a storage yard or factory.

Logistic robots

  • Logistic robots should be used in a limited area with dense building placement. In most cases this is the main factory area. With some updates the logistic bots can handle enormous amounts of items.
  • Bots are terrible at long distances and bulk goods: It's not a good idea use bots for ore-transport or from/to furnaces. It is possible to use them for this task, but don't hold your breath for this to work perfectly; there are several issues with the AI that cause them to work inefficiently more often than not. Some players take that problem as a challenge.
  • They do a perfect job on bulk-goods, if they can charge between their jobs, which is the case on train-stations: Train comes in, train is unloaded, bots transports to the target chests, bots charge. Next train comes in.
  • Bots do a perfect job if you use them for random transport of expensive products; transport intermediate products to where they're needed within the local network, especially if it's completely impractical to route a belt for the job.

Bots, while great at moving sparse items with slow production speeds, are a very costly thing to integrate with high-volume affairs like mining/smelting because they need frequent breaks to recharge.

The exception is the incredible benefit they offer to train stations because of the space-logistical challenge of routing and load-balancing belts around the train tracks, especially since it isn't necessarily 'constant flow', offering them time to recharge.

Seeing from throughput side

A (basic, yellow) transport belt cannot transport more than 800 items per minute and an express transport belt (the fastest) not more than 2400 (see above). Now compare that with a locomotive: a cargo wagon can deliver 2000 items (ores) per wagon! And if you smelt the resources at your outpost and transport intermediate plates (copper, iron) you can transport 4000 per wagon. So when speaking in terms of throughput, you need to swap to trains or make multiple parallel belts.

One might think this picture would change when you develop logistic robots. But that is not true! They can't be beaten for relatively small areas, especially with their researched bonuses, and they have some kind of intelligence; they try to deliver all requests as equally as possible. But bots are terrible at long distances and bulk goods. Here are some calculations, which explain the problem (all calculations are assumed with full upgrade for the logistic bots!):

Needed robots to transport 3000 items per minute over a distance of 50 tiles you need:

Robot speed:                     3 tiles/sec (basic speed) × 1.4 (logistic speed bonus) = 4.2 tiles/sec
Needed time for one transport:   100 tiles ÷ 4.2 tiles/sec = 24 secs
Need robots per minute:          24 secs ÷ 60 secs = you need 0.4 bots to transport one item per minute over 50 tiles.
Because of Logistic robot cargo: 0.4 bots ÷ 5 items per transport = 0.08 bots to transport one item per minute
Total:                           3000 items × 0.08 bots/per item = 240 bots

240 is a high, but still very realistic number for a logistic network with 4-5 roboports.

Needed robots to transport 3000 items per minute over a distance of 500 tiles you need:

Needed time for one transport:   1000 tiles ÷ 4.2 tiles/sec = 240 secs
Need robots per minute:          240 secs ÷ 60 secs = you need 4 bots to transport one item per minute over 500 tiles.
Because of logistic robot cargo: 4 bots ÷ 5 items per transport = 0.8 bots to transport one item per minute
Total:                           3000 items × 0.8 bots/per item = 2400 bots

2400 bots need a quite big amount of resources to produce, not accounting for the needed research (for bots, speed/stack size bonuses), number of roboports and the needed energy.

That leaves belts and trains.

Trains are faster than belts and offer better logistic control as you go between logistic networks.

Belts simply take forever, and you may wind up with a lot of asset tied up with simply being 'in transit'. You should see items sitting on a belt as "non-useable storage". Backed up belts can be a good thing for buffer space, but long belts are just inefficient, or better: You lose the control over the other end, due to the long gap; the resources on the other side could be long depleted, but you see it only five minutes later.

The effort to squeeze out the last 20% of belt performance increases depending on distance and total amount and at some point the train will win. But in your central factory area the bots will never be beaten. On the other hand: in small areas, and for simple throughput of one type of item (iron ore to iron-plates for example) we can place some belts in parallel, which is cheap and efficient (uses no energy) and when using parallel belts can achieve enormous throughput. For later game, the fast transport belt is a good compromise between cost and efficiency, and the Express transport belts should be use only for short sections.

The amount of robots you need is directly dependent on the density you build. The central part of your factory should be dense, because not only throughput is important here, but also speed: How fast can an item be produced in total (including the transportation time).

See also