There have been some comments recently about how AI-controlled ships in the player's fleet
appear to ignore orders, and how said orders
might mean something different to the AI than what the player thinks they mean.
This sounds like the sort of thing that can be tested, so I want to go ahead and do that. Let's explore how controllable the AI is, and what the 'best' combinations of personalities and orders are to achieve particular behavior.
Method:
- Define a
simple scenario where we want to see a certain behavior from a particular ship.
- Find a combination of personality and orders that achieves this behavior, if possible.
- Using FleetTester mod v1.0 and Starsector v0.95a-RC15.
Scenario #1a: Kiting a slower enemy with long-range weaponsSpoiler
Enemy: a lone Enforcer with 3x Light Autocannon, 2x Light Mortar, no missiles, and ITU. Steady personality.
Candidate ship: Hammerhead with 1x Hypervelocity Driver, 1x Heavy Mauler, 4x PD Laser, no missiles, and ITU.
>>> The Hammerhead outranges and outspeeds the Enforcer, which means it should be able to kite it without risk. Which combination of personality and orders achieves this?
> Steady without orders works.
> Steady + Engage works.
> Steady + Eliminate makes the Hammerhead close to minimum range.
> Steady + Search & Destroy works.
> Steady + Full Assault makes the Hammerhead close to minimum range.
> Aggressive without orders makes the Hammerhead close to minimum range.
> Aggressive + Engage, Eliminate, Search & Destroy OR Full Assault makes the Hammerhead close to minimum range.
> Aggressive + Avoid makes the Hammerhead run away like a scared chicken.
> Aggressive + Avoid + Search & Destroy still makes the Hammerhead run like a chicken.
> Cautious without orders works.
> Cautious + Engage works.
> Cautious + Eliminate makes the Hammerhead close to minimum range. Notably it retreats to vent flux quite soon as it builds up hard flux, something Steady didn't do nearly so much or so quickly.
> Cautious + Search & Destroy works.
> Cautious + Full Assault makes the Hammerhead close to Minimum range. Notably it does so more slowly than Steady, and also retreats much earlier as it builds up hard flux.
>>> Conclusion: Both Steady and Cautious AI are capable of kiting a slower enemy, keeping a steady pressure of weaponsfire until the target is destroyed. Both personalities can still be made to close the distance with specific orders. Aggressive AI is just too aggressive for kiting with long-range weapons, if the ship also has some shorter-range weapons like PD Lasers.
Scenario #1b: Kiting a slower enemy with a mix of short- and long-range weaponsSpoiler
Enemy: a lone Enforcer with 3x Light Autocannon, 2x Light Mortar and no other weapons. Steady personality.
Candidate ship: Hammerhead with 1x Hypervelocity Driver, 1x Heavy Mauler, 3x PD Laser and 1x Light Autocannon in one of the front smalls.
>>> The Hammerhead outspeeds the Enforcer, and outranges it with HVD/Mauler but also has a LAC with same range as the Enforcer's weapons. Will the AI opt to only use HVD/Mauler, or also try to use the LAC?
> No sense testing Aggressive: we already know it would close to PD Laser range in any case.
> Steady without orders makes the Hammerhead approach to LAC range.
> Steady + Engage OR Search & Destroy makes the Hammerhead approach to LAC range.
> Steady + Full Assault OR Eliminate makes the Hammerhead approach to PD Laser range.
> Steady + Avoid makes the Hammerhead run like a chicken.
> Cautious without orders works: the Hammerhead stays at HVD/Mauler range.
> Cautious + Engage OR Search & Destroy works as well.
> Cautious + Eliminate or Full Assault makes the Hammerhead approach to PD Laser range.
>>> Conclusion: When mixing weapons of different ranges, Steady AI chooses engagement range based on its
shortest-range (non-PD) weapons, while Cautious AI chooses engagement ranged based on its
longest-range (non-PD) weapons.
Scenario #2: Capturing and controlling strategic points / Rallying at a waypointSpoiler
Enemy: 2x Kite (A) with a Light Dual Autocannon, ITU, and no other weapons. Steady personality.
Candidate ship: Eagle with 3x Railgun, 3x Pulse Laser and 5x PD Laser, ITU, and no missiles.
Mission: capture and control/defend a Comm Relay in the center of the battlefield.
>>> Two lightly armed Kites are no match for this Eagle, but they are a lot faster. Is there a way to make the Eagle stay near the strategic point and hold it, or will it inevitably be distracted by a Kite? For reference: manually piloting the Eagle this mission is a breeze. Park the Eagle on the Comm Relay, set all weapons to autofire and just turn the ship with Maneuvering Jets towards the Kites. Easy.
> Steady + Capture + Control makes the Eagle get distracted by one Kite long enough for the other Kite to capture the Comm Relay. Subsequently the Eagle does re-capture it. The same pattern then repeats. Sometimes the Eagle even prioritizes fighting the Kites over re-capturing the Comm Relay.
> Steady + Assault + Defend still makes the Eagle get distracted long enough to lose the Comm Relay.
> Steady + Rally Task Force order on the Comm Relay (instead of Capture/Assault) is no different.
Observation: Enemy Kites often prioritize flanking the Eagle to shoot at its relatively unprotected rear over capturing the Comm Relay, even when the Eagle has wandered far enough away that one of the Kites could easily capture it.
> Aggressive + Capture + Control also makes the Eagle prone to getting distracted and losing the Comm Relay.
> Aggressive + Assault + Defend is no different.
> Aggressive + Rally Task Force order on the Comm Relay is no different.
> Cautious + Capture + Control is no different.
> Cautious + Assault + Defend is no different.
> Cautious + Rally Task Force on the Comm Relay is no different.
> Timid personality makes the Eagle actively back away from the Kites even if that means completely ignoring a Control, Defend or Rally order.
>>> Conclusion: Ship AI appears to prioritize engaging enemies over staying at an assigned strategic point or rally point. And there is no way for the player to override that behavior. Which is kind of disappointing.
More scenario's to follow in the future. No promises on how far in the future that will be. Leave your suggestions for scenarios to explore!