Several activities that queue child Move activities can get into a bad scenario where the actor is pathfinding and then gets stuck because the destination is unreachable. When the Move activity then completes, then parent activity sees it has yet to reach the destination and tries to move again. However, the actor is still blocked in the same spot as before and thus the movment finishes immediately. This causes a performance death spiral where the actor attempts to pathfind every tick. The pathfinding attempt can also be very expensive if it must exhaustively check the whole map to determine no route is possible.
In order to prevent blocked actors from running into this scenario, we introduce MoveCooldownHelper. In its default setup it allows the parent activity to bail out if the actor was blocked during a pathfinding attempt. This means the activity will be dropped rather than trying to move endlessly. It also has an option to allow retrying if pathfinding was blocked, but applies a cooldown to avoid the performance penalty. For activities such as Enter, this means the actors will still try and enter their target if it is unreachable, but will only attempt once a second now rather than every tick.
MoveAdjacentTo will now cancel if it fails to reach the destination. This fixes MoveOntoAndTurn to skip the Turn if the move didn't reach the intended destination. Any other derived classes will similarly benefit from skipping follow-up actions.
This rule no longer appears to be buggy, so enforce it. Some of the automated fixes are adjusted in order to improve the result. #pragma directives have no option to control indentation, so remove them where possible.
Previously, the ClosestTo and PositionClosestTo existed to perform a simple distance based check to choose the closest location from a choice of locations to a single other location. For some functions this is sufficient, but for many functions we want to then move between the locations. If the location selected is in fact unreachable (e.g. on another island) then we would not want to consider it.
We now introduce ClosestToIgnoringPath for checks where we don't care about a path existing, e.g. weapons hitting nearby targets. When we do care about paths, we introduce ClosestToWithPathFrom and ClosestToWithPathTo which will check that a path exists. The PathFrom check will make sure one of the actors from the list can make it to the single target location. The PathTo check will make sure the single actor can make it to one of the target locations. This difference allows us to specify which actor will be doing the moving. This is important as a path might exists for one actor, but not another. Consider two islands with a hovercraft on one and a tank on the other. The hovercraft can path to the tank, but the tank cannot path to the hovercraft.
We also introduce WithPathFrom and WithPathTo. These will perform filtering by checking for valid paths, but won't select the closest location.
By employing the new methods that filter for paths, we fix various behaviour that would cause actors to get confused. Imagine an islands map, by checking for paths we ensure logic will locate reachable locations on the island, rather than considering a location on a nearby island that is physically closer but unreachable. This fixes AI squad automation, and other automated behaviours such as rearming.
The Harvester trait and MoveAdjacentTo activity called the pathfinder but had a single source and multiple targets. The pathfinder interface only allows for the opposite: multiple sources and a single target. To work around this they would swap the inputs. This works in most cases but not all cases. One aspect of asymmetry is that an actor may move out of an inaccessible source cell, but not onto an inaccessible target cell.
Searches that involved an inaccessible source cell and that applied this swapping method would therefore fail to return a path, when a valid path was possible. Although a rare case, once good way to reproduce is to use a production building that spawns actors on inaccessible cells around it, such as the RA naval yard. A move order uses the pathfinder correctly and the unit will move out. Using a force attack causes the unit to use the broken "swapped" mechanism in MoveAdjacentTo and it will be stuck.
This asymmetry has been longstanding but the pathfinding infrastructure only sporadically accounted for it. It is now documented and applied consistently. Create a new overload on the pathfinder trait that allows a single source and multiple targets, so callers have an overload that does what they need and won't be tempted to swap the positions and run into this issue.
Internally, this requires us to teach Locomotor to ignore the self actor when performing movement cost checks for these "in reverse" searches so the unit doesn't consider the cell blocked by itself.
Replaces the existing bi-directional search between points used by the pathfinder with a guided hierarchical search. The old search was a standard A* search with a heuristic of advancing in straight line towards the target. This heuristic performs well if a mostly direct path to the target exists, it performs poorly it the path has to navigate around blockages in the terrain. The hierarchical path finder maintains a simplified, abstract graph. When a path search is performed it uses this abstract graph to inform the heuristic. Instead of moving blindly towards the target, it will instead steer around major obstacles, almost as if it had been provided a map which ensures it can move in roughly the right direction. This allows it to explore less of the area overall, improving performance.
When a path needs to steer around terrain on the map, the hierarchical path finder is able to greatly improve on the previous performance. When a path is able to proceed in a straight line, no performance benefit will be seen. If the path needs to steer around actors on the map instead of terrain (e.g. trees, buildings, units) then the same poor pathfinding performance as before will be observed.
Some path searches, using PathSearch, were created directly at the callsite rather than using the pathfinder trait. This means some searches did not not benefit from the performance checks done in the pathfinder trait. It also means the pathfinder trait was not responsible for all pathing done in the game. Fix this with the following changes:
- Create a sensible shape for the IPathFinder interface and promote it to a trait interface, allowing theoretical replacements of the implementation. Ensure none of the concrete classes in OpenRA.Mods.Common.Pathfinder are exposed in the interface to ensure this is possible.
- Update the PathFinder class to implement the interface, and update several callsites manually running pathfinding code to instead call the IPathFinder interface.
- Overall, this allows any implementation of the IPathFinder interface to intercept and control all path searching performed by the game. Previously some searches would not have used it, and no alternate implementations were possible as the existing implementation was hardcoded into the interface shape.
Additionally:
- Move the responsibility of finding paths on completed path searches from pathfinder to path search, which is a more sensible location.
- Clean up the pathfinder pre-search optimizations.
- When a path search is being performed the path search will not attempt route to inaccessible cells, so domain index checks to avoid inaccessible cells in the search predicate are redundant and can be removed.
- DomainIndex is a required world trait, so we don't need to use TraitOrDefault and therefore can avoid dealing with the null case.
The existing APIs surfaces for pathfinding are in a wonky shape. We rearrange various responsibilities to better locations and simplify some abstractions that aren't providing value.
- IPathSearch, BasePathSearch and PathSearch are combined into only PathSearch. Its role is now to run a search space over a graph, maintaining the open queue and evaluating the provided heuristic function. The builder-like methods (WithHeuristic, Reverse, FromPoint, etc) are removed in favour of optional parameters in static creation methods. This removes confusion between the builder-aspect and the search function itself. It also becomes responsible for applying the heuristic weight to the heuristic. This fixes an issue where an externally provided heuristic ignored the weighting adjustment, as previously the weight was baked into the default heuristic only.
- Reduce the IGraph interface to the concepts of nodes and edges. Make it non-generic as it is specifically for pathfinding, and rename to IPathGraph accordingly. This is sufficient for a PathSearch to perform a search over any given IGraph. The various customization options are concrete properties of PathGraph only.
- PathFinder does not need to deal with disposal of the search/graph, that is the caller's responsibility.
- Remove CustomBlock from PathGraph as it was unused.
- Remove FindUnitPathToRange as it was unused.
- Use PathFinder.NoPath as the single helper to represent no/empty paths.
Each successive value of BlockedByActor is a superset of the previous value. Having a mixed up order of values in PathSearchOrder is not useful.
In the previous ordering, if a search for Immovable failed to find a path, it would then attempt Stationary. However Stationary is *more* restrictive then Immovable. If Immovable failed, there is no way Stationary could succeed. This means the search for Stationary is wasted effort.
In the fixed ordering, we try Stationary first. In the fixed ordering there are no pointless searches. Every search might succeed where the previous one failed and is therefore useful to try.
If progress == Distance, we must not move again on the same tick,
but still 'return true' to avoid losing a tick in the case
when this is the last Move tick followed by a different activity
(or a new queued Move, for example via waypoints).
There were 2 issues at work here, both when progress
would overshoot Distance (which is the usual case,
rather than the exception):
- the overshot progress was passed on by MoveFirstHalf, however
OnComplete would make the next MovePart start ticking the
same tick on which the old MovePart reached Distance,
but move by carryoverProgress +(!!!) terrain speed instead of moving
by just the left-over carryoverProgress.
- MoveSecondHalf would not pass any overshot progress to the
next MoveFirstHalf queued by parent Move, leading to
the next MoveFirstHalf performing a full-speed move the same tick
MoveSecondHalf finished its last move.
InnerActivity and UpdateCenterLocation made this
overly complex and hard to read & debug.
This also fixes a bug that would make an outdated
facing being passed during OnComplete (because
InnerActivity was cached before UpdateCenterLocation
could set the correct final facing).
