def init_variables(self, info):
# Here you have the information of the game (virtual init() in random_walk.cpp)
# List: game_time, goal, number_of_robots, penalty_area, codewords,
# robot_height, robot_radius, max_linear_velocity, field, team_info,
# {rating, name}, axle_length, resolution, ball_radius
# self.game_time = info['game_time']
# self.field = info['field']
self.max_linear_velocity = info['max_linear_velocity']
self.resolution = info['resolution']
self.colorChannels = 3 # nf
self.end_of_frame = False
self.image = Received_Image(self.resolution, self.colorChannels)
self.D = [] # Replay Memory
self.update = 100 # Update Target Network
self.epsilon = 1.0 # Initial epsilon value
self.final_epsilon = 0.05 # Final epsilon value
self.dec_epsilon = 0.05 # Decrease rate of epsilon for every generation
self.step_epsilon = 20000 # Number of iterations for every generation
self.observation_steps = 5000 # Number of iterations to observe before training every generation
self.save_every_steps = 5000 # Save checkpoint
self.num_actions = 11 # Number of possible possible actions
self._frame = 0
self._iterations = 0
self.minibatch_size = 64
self.gamma = 0.99
self.sqerror = 100 # Initial sqerror value
self.Q = NeuralNetwork(
None, False, False
) # 2nd term: False to start training from scratch, use CHECKPOINT to load a checkpoint
self.Q_ = NeuralNetwork(self.Q, False, True)
self.wheels = [0 for _ in range(10)]
return
def init_variables(self, info):
# Here you have the information of the game (virtual init() in random_walk.cpp)
# List: game_time, number_of_robots
# field, goal, penalty_area, goal_area, resolution Dimension: [x, y]
# ball_radius, ball_mass,
# robot_size, robot_height, axle_length, robot_body_mass, ID: [0, 1, 2, 3, 4]
# wheel_radius, wheel_mass, ID: [0, 1, 2, 3, 4]
# max_linear_velocity, max_torque, codewords, ID: [0, 1, 2, 3, 4]
# self.game_time = info['game_time']
# self.number_of_robots = info['number_of_robots']
# self.field = info['field']
# self.goal = info['goal']
# self.penalty_area = info['penalty_area']
# self.goal_area = info['goal_area']
self.resolution = info['resolution']
# self.ball_radius = info['ball_radius']
# self.ball_mass = info['ball_mass']
# self.robot_size = info['robot_size']
# self.robot_height = info['robot_height']
# self.axle_length = info['axle_length']
# self.robot_body_mass = info['robot_body_mass']
# self.wheel_radius = info['wheel_radius']
# self.wheel_mass = info['wheel_mass']
self.max_linear_velocity = info['max_linear_velocity']
# self.max_torque = info['max_torque']
# self.codewords = info['codewords']
self.total_distance=0
self.ball_touch=0
self.colorChannels = 3 # nf
self.end_of_frame = False
self.image = Received_Image(self.resolution, self.colorChannels)
self.received_frame = Frame()
self.D = [] # Replay Memory
self.distance_buffer=[] #distance buffer for reward
self.update = 100 # Update Target Network
self.epsilon = 1.0 # Initial epsilon value
self.final_epsilon = 0.05 # Final epsilon value
self.dec_epsilon = 0.05 # Decrease rate of epsilon for every generation
self.step_epsilon = 5000 # Number of iterations for every generation
self.observation_steps = 1000 # Number of iterations to observe before training every generation
self.save_every_steps = 1000 # Save checkpoint
self.num_actions = 11 # Number of possible possible actions
self._frame = 0
self._iterations = 0
self.minibatch_size = 64
self.gamma = 0.99
self.sqerror = 100 # Initial sqerror value
self.Q = NeuralNetwork(None, False, False) # 2nd term: False to start training from scratch, use CHECKPOINT to load a checkpoint
self.Q_ = NeuralNetwork(self.Q, False, True)
self.wheels = [0 for _ in range(10)]
return
def init_variables(self, info):
# Here you have the information of the game (virtual init() in random_walk.cpp)
# List: game_time, goal, number_of_robots, penalty_area, codewords,
# robot_height, robot radius, max_linear_velocity, field, team_info,
# {rating, name}, axle_length, resolution, ball_radius
# self.game_time = info['game_time']
# self.field = info['field']
self.max_linear_velocity = info['max_linear_velocity']
self.resolution = info['resolution']
self.colorChannels = 3 # nf in dqn_main.py
self.end_of_frame = False
self.image = Received_Image(self.resolution, self.colorChannels)
self._frame = 0
self.Q = NeuralNetwork(None, CHECKPOINT, False) # 2nd term: False to start training from scratch, use CHECKPOINT to load a checkpoint
self.wheels = [0 for _ in range(10)]
return
def init_variables(self, info):
# Here you have the information of the game (virtual init() in random_walk.cpp)
# List: game_time, number_of_robots
# field, goal, penalty_area, goal_area, resolution Dimension: [x, y]
# ball_radius, ball_mass,
# robot_size, robot_height, axle_length, robot_body_mass, ID: [0, 1, 2, 3, 4]
# wheel_radius, wheel_mass, ID: [0, 1, 2, 3, 4]
# max_linear_velocity, max_torque, codewords, ID: [0, 1, 2, 3, 4]
# self.game_time = info['game_time']
# self.number_of_robots = info['number_of_robots']
# self.field = info['field']
# self.goal = info['goal']
# self.penalty_area = info['penalty_area']
# self.goal_area = info['goal_area']
self.resolution = info['resolution']
# self.ball_radius = info['ball_radius']
# self.ball_mass = info['ball_mass']
# self.robot_size = info['robot_size']
# self.robot_height = info['robot_height']
# self.axle_length = info['axle_length']
# self.robot_body_mass = info['robot_body_mass']
# self.wheel_radius = info['wheel_radius']
# self.wheel_mass = info['wheel_mass']
self.max_linear_velocity = info['max_linear_velocity']
# self.max_torque = info['max_torque']
# self.codewords = info['codewords']
self.colorChannels = 3 # nf in dqn_main.py
self.end_of_frame = False
self.image = Received_Image(self.resolution, self.colorChannels)
self._frame = 0
self.Q = NeuralNetwork(
None, CHECKPOINT, False
) # 2nd term: False to start training from scratch, use CHECKPOINT to load a checkpoint
self.wheels = [0 for _ in range(10)]
return