Update rewared structure (fixed major bugs)

configs/game/monster_config.py

@@ -6,7 +6,7 @@ asset_path = os.path.join(
     script_dir, '../..', 'assets')
 
 monster_data = {
-    'squid': {'id': 1, 'health': 100, 'exp': 100, 'attack': 20, 'attack_type': 'slash', 'attack_sound': f'{asset_path}/audio/attack/slash.wav', 'speed': 3, 'knockback': 20, 'attack_radius': 80, 'notice_radius': 360},
-    'raccoon': {'id': 2, 'health': 300, 'exp': 250, 'attack': 40, 'attack_type': 'claw',  'attack_sound': f'{asset_path}/audio/attack/claw.wav', 'speed': 2, 'knockback': 20, 'attack_radius': 120, 'notice_radius': 400},
-    'spirit': {'id': 3, 'health': 100, 'exp': 110, 'attack': 8, 'attack_type': 'thunder', 'attack_sound': f'{asset_path}/audio/attack/fireball.wav', 'speed': 4, 'knockback': 20, 'attack_radius': 60, 'notice_radius': 350},
-    'bamboo': {'id': 4, 'health': 70, 'exp': 120, 'attack': 6, 'attack_type': 'leaf_attack', 'attack_sound': f'{asset_path}/audio/attack/slash.wav', 'speed': 3, 'knockback': 20, 'attack_radius': 50, 'notice_radius': 300}}
+    'squid': {'id': 1, 'health': .1, 'exp': 1, 'attack': .5, 'attack_type': 'slash', 'speed': 3, 'knockback': 20, 'attack_radius': 80, 'notice_radius': 360},
+    'raccoon': {'id': 2, 'health': .3, 'exp': 2.5, 'attack': .8, 'attack_type': 'claw',  'speed': 2, 'knockback': 20, 'attack_radius': 120, 'notice_radius': 400},
+    'spirit': {'id': 3, 'health': .1, 'exp': 1.1, 'attack': .6, 'attack_type': 'thunder', 'speed': 4, 'knockback': 20, 'attack_radius': 60, 'notice_radius': 350},
+    'bamboo': {'id': 4, 'health': .07, 'exp': 1.2, 'attack': .4, 'attack_type': 'leaf_attack', 'speed': 3, 'knockback': 20, 'attack_radius': 50, 'notice_radius': 300}}

configs/game/player_config.py

@@ -1,9 +1,9 @@
 tank_stats = {
     'role_id': 1,
-    'health': 150,
-    'energy': 40,
-    'attack': 7,
-    'magic': 3,
+    'health': 1.5,
+    'energy': .4,
+    'attack': .7,
+    'magic': .3,
     'speed': 3
 }
 

configs/game/spell_config.py

@@ -5,5 +5,5 @@ asset_path = os.path.join(
     script_dir, '../..', 'assets')
 
 magic_data = {
-    'flame': {'strength': 5, 'cost': 20, 'graphic': f"{asset_path}/graphics/particles/flame/fire.png"},
-    'heal': {'strength': 20, 'cost': 10, 'graphic': f"{asset_path}/graphics/particles/heal/heal.png"}}
+    'flame': {'strength': 5, 'cost': .020, 'graphic': f"{asset_path}/graphics/particles/flame/fire.png"},
+    'heal': {'strength': 20, 'cost': .010, 'graphic': f"{asset_path}/graphics/particles/heal/heal.png"}}

entities/components/stats.py

@@ -42,6 +42,12 @@ class StatsHandler:
         else:
             self.energy = self.stats['energy']
 
+    def health_recovery(self):
+        if self.energy < self.stats['health']:
+            self.energy += 0.15
+        else:
+            self.energy = self.stats['energy']
+
     def get_value_by_index(self, index):
         return list(self.stats.values())[index]
 

entities/player.py

@@ -16,7 +16,16 @@ from agents.ppo.agent import Agent
 
 class Player(pygame.sprite.Sprite):
 
-    def __init__(self, position, groups, obstacle_sprites, visible_sprites, attack_sprites, attackable_sprites, role, player_id):
+    def __init__(self,
+                 position,
+                 groups,
+                 obstacle_sprites,
+                 visible_sprites,
+                 attack_sprites,
+                 attackable_sprites,
+                 role,
+                 player_id):
+
         super().__init__(groups)
 
         # Setup Sprites
@@ -79,7 +88,9 @@ class Player(pygame.sprite.Sprite):
                             offset = pygame.math.Vector2(0, 75)
                             for leaf in range(randint(3, 6)):
                                 self.animation_player.create_grass_particles(
-                                    position=pos - offset, groups=[self.visible_sprites])
+                                    position=pos - offset,
+                                    groups=[self.visible_sprites])
+
                             target_sprite.kill()
                         else:
                             target_sprite.get_damaged(
@@ -109,13 +120,12 @@ class Player(pygame.sprite.Sprite):
 
         self.reward_features = [
             self.stats.exp,
-            # nearest_dist,
-            -nearest_enemy.stats.health,
-            self.stats.health
+            np.exp(-nearest_dist**2),
+            np.exp(-nearest_enemy.stats.health**2),
+            -np.exp(-self.stats.health)
         ]
 
         self.state_features = [
-            # TODO: Find a way to not use magic numbers
             np.exp(-self.rect.center[0]),
             np.exp(-self.rect.center[1]),
             self._input.movement.direction.x,
@@ -127,7 +137,6 @@ class Player(pygame.sprite.Sprite):
         enemy_states = []
 
         for distance, direction, enemy in sorted_distances[:5]:
-            # TODO: Find a way to not use magic numbers
             enemy_states.extend([
                 np.exp(-distance),
                 direction[0],
@@ -156,12 +165,13 @@ class Player(pygame.sprite.Sprite):
             n_epochs=4)
         try:
             self.agent.load_models()
-        except FileNotFoundError as e:
-            print(f"{e}. Skipping loading...")
+        except FileNotFoundError:
+            print("FileNotFoundError for agent.load_model().\
+                Skipping loading...")
 
     def is_dead(self):
-        if self.stats.health == 0:
-            self.stats.exp = -100
+        if self.stats.health <= 0:
+            self.stats.exp = max(0, self.stats.exp - .5)
             return True
         else:
             return False
@@ -182,11 +192,9 @@ class Player(pygame.sprite.Sprite):
                                 self.animation.rect,
                                 self)
 
-        self.done = self.is_dead()
-
         self.score = self.stats.exp
         self.agent.remember(self.state_features, action,
-                            probs, value, self.stats.exp, self.done)
+                            probs, value, self.stats.exp, self.is_dead())
 
         if self.n_steps % self.N == 0:
             self.agent.learn()
@@ -194,7 +202,7 @@ class Player(pygame.sprite.Sprite):
 
         self.get_current_state()
 
-        if self.done:
+        if self.is_dead():
             self.agent.learn()
 
         # Refresh objects based on input
@@ -207,5 +215,6 @@ class Player(pygame.sprite.Sprite):
         self.rect = self.animation.rect
 
         # Cooldowns and Regen
+        self.stats.health_recovery()
         self.stats.energy_recovery()
         self._input.cooldowns(self._input.combat.vulnerable)

game.py

@@ -13,17 +13,26 @@ class Game:
 
         self.screen = pygame.display.set_mode(
             (WIDTH, HEIGHT))
+
         pygame.display.set_caption('Pneuma')
+
+        img = pygame.image.load('assets/graphics/icon.png')
+        pygame.display.set_icon(img)
         self.clock = pygame.time.Clock()
 
         self.level = Level()
 
+    def calc_score(self):
+        self.scores = [0 for _ in range(len(self.level.player_sprites))]
+
+        for player in self.level.player_sprites:
+            self.scores[player.player_id] = player.stats.exp
+
     def run(self):
 
         for event in pygame.event.get():
             if event.type == pygame.QUIT:
-                pygame.quit()
-                sys.exit()
+                self.quit()
             if event.type == pygame.KEYDOWN:
                 if event.key == pygame.K_m:
                     self.level.toggle_menu()
@@ -34,3 +43,7 @@ class Game:
 
         pygame.display.update()
         self.clock.tick(FPS)
+
+    def quit(self):
+        pygame.quit()
+        sys.exit()

level/level.py

@@ -1,5 +1,6 @@
 import os
 import pygame
+import numpy as np
 
 from random import choice
 
@@ -46,6 +47,7 @@ class Level:
         else:
             for player in self.player_sprites:
                 player.get_max_num_states()
+        self.dead_players = np.zeros(len(self.player_sprites))
 
         # UI setup
         self.ui = UI()
@@ -174,11 +176,13 @@ class Level:
             self.visible_sprites.custom_draw(self.player)
             self.ui.display(self.player)
 
-        debug('v0.6')
+        debug('v0.8')
 
         for player in self.player_sprites:
             if player.is_dead():
-                self.done = True
+                self.dead_players[player.player_id] = True
+
+        self.done = True if self.dead_players.all() == 1 else False
 
         if not self.game_paused:
             # Update the game

main.py

@@ -1,3 +1,6 @@
+import numpy as np
+import matplotlib.pyplot as plt
+
 from game import Game
 from tqdm import tqdm
 
@@ -6,54 +9,67 @@ environ['PYGAME_HIDE_SUPPORT_PROMPT'] = '1'
 
 
 if __name__ == '__main__':
-    n_episodes = 3000
+
+    n_episodes = 1000
+    game_len = 10000
 
     figure_file = 'plots/score.png'
-    score_history = []
     best_score = 0
     avg_score = 0
 
-    agent_list = []
-
-    game_len = 5000
-
     game = Game()
 
+    agent_list = []
+    exp_points_list = []
+    score_history = np.zeros(
+        shape=(len(game.level.player_sprites), n_episodes, ))
+    best_score = np.zeros(len(game.level.player_sprites))
+    avg_score = np.zeros(len(game.level.player_sprites))
     for i in tqdm(range(n_episodes)):
         # TODO: Make game.level.reset_map() so we don't __init__ everything all the time (such a waste)
         if i != 0:
             game.level.__init__(reset=True)
         # TODO: Make game.level.reset_map() so we don't pull out and load the agent every time (There is -definitevly- a better way)
             for player in game.level.player_sprites:
-            for player_id, agent in agent_list:
-                if player.player_id == player_id:
-                    player.agent = agent
+                for agent in agent_list:
+                    player.agent = agent_list[player.player_id]
+                    player.stats.exp = score_history[player.player_id][i-1]
 
         agent_list = []
-        done = False
-        score = 0
-        for _ in tqdm(range(game_len)):
-            if not game.level.done:
-                game.run()
-            else:
-                break
-        for player in game.level.player_sprites:
-            agent_list.append((player.player_id, player.agent))
 
-        if i == n_episodes-1 and game.level.enemy_sprites != []:
+        for j in range(game_len):
+            if not game.level.done:
+
+                game.run()
+                game.calc_score()
+
+                if (j == game_len-1 or game.level.done) and game.level.enemy_sprites != []:
                     for player in game.level.player_sprites:
                         for enemy in game.level.enemy_sprites:
-                    player.stats.exp -= 5
-                player.update()
+                            player.stats.exp *= .95
+            else:
+                break
 
         for player in game.level.player_sprites:
+            agent_list.append(player.agent)
+            exp_points = player.stats.exp
+            score_history[player.player_id][i] = exp_points
+            avg_score[player.player_id] = np.mean(
+                score_history[player.player_id])
+            if avg_score[player.player_id] >= best_score[player.player_id]:
                 player.agent.save_models()
+                best_score[player.player_id] = avg_score[player.player_id]
 
-        # TODO: Make it so that scores appear here for each player
-        # score_history.append(game.level.player.score)
-        # print(score)
-        # avg_score = np.mean(score_history[-100:])
+            print(
+                f"\nCumulative score for player {player.player_id}:\
+                    {score_history[0][i]}\
+                    \nAverage score for player {player.player_id}:\
+                    {avg_score[player.player_id]}\
+                    \nBest score for player {player.player_id}:\
+                    {best_score[player.player_id]}")
 
-        # if avg_score > best_score:
-        #     best_score = avg_score
-        #     game.level.player.agent.save_models()
+    plt.plot(score_history[0])
+
+    game.quit()
+
+    plt.show()