docs: update docs

README.md

@@ -4,17 +4,20 @@ A sokoban with a solver.
 
 ## Features
 
+- Integrated solver.
 - Full mouse control.
+- Player animation.
+- Front-end and back-end separation.
+- Levels and solutions are stored in the database.
+
+### [Auto crate push](docs/auto_crate_push.md)
 
 <p align="center"><img src="./docs/auto_crate_push.gif" width=50%></p>
 
-- Integrated solver.
+### [Solver](docs/solver.md)
 
 <p align="center"><img src="./docs/auto_solve.gif" width=50%></p>
 
-- Front-end and back-end separation.
-- Levels and solutions are stored in the database.
-
 ## Keymap
 
 ### Mouse

docs/auto_crate_push.md

@@ -0,0 +1,30 @@
+# Auto crate push
+
+<p align="center"><img src="auto_crate_push.gif" width=70%></p>
+
+This feature can provide good assistance when users play larger levels.  
+At the same time, it also retains the traditional control method.
+
+Clicking on a crate will display all the points that the crate can reach (without moving other crates).
+
+Take Microban #155 as an example:
+
+<p align="center"><img src="auto_crate_push_1.png" width=70%></p>
+
+Clicking on one of the points will automatically push the crate to that position.
+
+For example, user can click on the target, and the character will automatically push the crate to the target to complete the level.
+
+Some areas where the cratees are reachable do not display points. This is because pushing the crate to those locations will lead to a dead end and the player will not be able to continue to complete the level.
+
+<p align="center"><img src="auto_crate_push_2.png" width=70%></p>
+
+## Controversial
+
+This feature is a bit controversial, with some people saying it's akin to cheating.
+
+For simple levels, this does significantly reduce the level difficulty. An extreme example is a level with only a single crate and target, which means the player can complete it without having to do any reasoning.
+
+But for complex levels, the difficulty mainly lies in the complex pushing relationship between multiple cratees, rather than the pushing of a single crate. This feature allows players to focus on more complex reasoning instead of repeating the simple work of pushing a single crate.
+
+Otherwise, this feature is **optional**. Users can always use traditional control methods.

docs/solver.md

@@ -0,0 +1,34 @@
+# Solver
+
+<p align="center"><img src="auto_solve.gif" width=70%></p>
+
+The solver can automatically solve simple levels.
+
+## Strategy
+
+- `Fast`: Speed priority.
+- `Mixed`: Balanced speed and steps.
+- `OptimalMovePush`: Find move optimal solutions with best pushes.
+- `OptimalPushMove`: Find push optimal solutions with best moves.
+
+## Lower bound calculation method
+
+- `PushCount`: Minimum push count to nearest target.
+- `MoveCount`: Minimum move count to nearest target.
+- `ManhattanDistance`: Manhattan distance to nearest target.
+
+## Optimization
+
+- Deadlocks detection.
+  - Dead square deadlocks.
+  - Freeze deadlocks.
+- Tunnels detection.
+
+## Visualization
+
+Supports visualization of the automatic solution process. This feature can be used to intuitively view the working status of the solver.
+
+- Displays the optimal state obtained by the current solver.
+- Display the lower bound as a heat map.
+
+![Visualization](solver_visualization.png)