dribble

.idea/FCPCodebase.iml

@@ -2,7 +2,7 @@
 <module type="PYTHON_MODULE" version="4">
   <component name="NewModuleRootManager">
     <content url="file://$MODULE_DIR$" />
-    <orderEntry type="jdk" jdkName="FCP" jdkType="Python SDK" />
+    <orderEntry type="jdk" jdkName="FCPCodebase" jdkType="Python SDK" />
     <orderEntry type="sourceFolder" forTests="false" />
   </component>
   <component name="PyDocumentationSettings">

.idea/misc.xml

@@ -3,5 +3,5 @@
   <component name="Black">
     <option name="sdkName" value="fcp_env" />
   </component>
-  <component name="ProjectRootManager" version="2" project-jdk-name="FCP" project-jdk-type="Python SDK" />
+  <component name="ProjectRootManager" version="2" project-jdk-name="FCPCodebase" project-jdk-type="Python SDK" />
 </project>

agent/Agent.py

@@ -32,6 +32,8 @@ class Agent(Base_Agent):
             distance_to_final_target = np.linalg.norm(target_2d - self.loc_head_position[:2])
         return self.behavior.execute("Dribble", next_rel_ori,
                                      False)  # Args: target, is_target_abs, ori, is_ori_abs, distance
+    def push(self, target_2d=(15, 0), avoid_obstacles=True):
+        self.behavior.execute("Push_RL")
 
     def beam(self, avoid_center_circle=False):
         r = self.world.robot

behaviors/Behavior.py

@@ -49,7 +49,9 @@ class Behavior():
         from behaviors.custom.Get_Up.Get_Up import Get_Up
         from behaviors.custom.Step.Step import Step
         from behaviors.custom.Walk.Walk import Walk
-        classes = [Basic_Kick,Dribble,Fall,Get_Up,Step,Walk]
+        from behaviors.custom.Push_RL.Push_RL import Push_RL
+
+        classes = [Basic_Kick,Dribble,Fall,Get_Up,Step,Walk,Push_RL]
 
         '''---- End of manual declarations ----'''
 

behaviors/custom/Dribble/Dribble.py

@@ -7,7 +7,6 @@ import pickle
 
 
 class Dribble():
-
     def __init__(self, base_agent : Base_Agent) -> None:
         self.behavior = base_agent.behavior
         self.path_manager = base_agent.path_manager

behaviors/custom/Dribble/Env.py

@@ -178,4 +178,4 @@ class Env():
         self.execute_ik(l_ankle_pos, l_foot_rot, r_ankle_pos, r_foot_rot)           # legs 
         r.set_joints_target_position_direct( slice(14,22), arms, harmonize=False )  # arms
 
-        self.step_counter += 1
+        self.step_counter += 1

behaviors/custom/Push_RL/Env_HL.py

@@ -0,0 +1,79 @@
+import math
+from typing import List
+import numpy as np
+from math_ops.Math_Ops import Math_Ops as U
+from behaviors.custom.Step.Step_Generator import Step_Generator
+from world.commons import Other_Robot
+from world.World import World
+from agent.Base_Agent import Base_Agent
+
+
+class Env_HL:
+    COLS = 16
+    LINS = 5
+
+    def __init__(self, base_agent: Base_Agent):
+        self.world = base_agent.world
+        self.obs = np.zeros(163, np.float32)
+
+        self.output = 0
+
+    def fill_radar(self, radar, team=None, radar_part=None, RADIAL_START=None, RADIAL_MULT=None):
+        if RADIAL_MULT is None:
+            RADIAL_MULT = {'team': List[Other_Robot]}
+        w = self.world
+        bp = w.ball_abs_pos[:2]
+        C = self.COLS
+        L = self.LINS
+        vec_b_goal = (15.5, 0) - bp
+        vec_b_goal_absdir = U.vector_angle(vec_b_goal)
+        dist_closest_player = 10
+        for t in team:
+            if w.time_local_ms - t.state_last_update > 500:
+                continue
+            vec_b_opp = t.state_abs_pos[:2] - bp
+            dist_b_opp = np.linalg.norm(vec_b_opp)
+            if dist_b_opp < dist_closest_player:
+                dist_closest_player = dist_b_opp
+            vec_b_opp_dir = U.normalize_deg(U.vector_angle(vec_b_opp) - vec_b_goal_absdir)
+            (div, mod) = divmod(vec_b_opp_dir + 180, 360 / C)
+            zone = int(div) % C
+            prog = mod * C / 360
+            ang_column_weight_1 = (zone, 1 - prog)
+            ang_column_weight_2 = ((zone + 1) % C, prog)
+            zone = max(1, 1 + math.log((dist_b_opp + 1e-06) / RADIAL_START, RADIAL_MULT))
+            prog = zone % 1
+            zone = math.ceil(zone) - 1
+            if zone >= L:
+                continue
+            rad_line_weight_1 = None if zone == 0 else (zone - 1, 1 - prog)
+            rad_line_weight_2 = (zone, 1 if zone == 0 else prog)
+            if rad_line_weight_1 is not None:
+                radar[(radar_part, rad_line_weight_1[0], ang_column_weight_1[0])] += rad_line_weight_1[1] * \
+                                                                                     ang_column_weight_1[1]
+                radar[(radar_part, rad_line_weight_1[0], ang_column_weight_2[0])] += rad_line_weight_1[1] * \
+                                                                                     ang_column_weight_2[1]
+            radar[(radar_part, rad_line_weight_2[0], ang_column_weight_1[0])] += rad_line_weight_2[1] * \
+                                                                                 ang_column_weight_1[1]
+            radar[(radar_part, rad_line_weight_2[0], ang_column_weight_2[0])] += rad_line_weight_2[1] * \
+                                                                                 ang_column_weight_2[1]
+        return dist_closest_player
+
+    def observe(self, init=False):
+        if init:
+            self.output = 0
+        radar = np.zeros((2, self.LINS, self.COLS))
+        RADIAL_START = 0.3
+        RADIAL_MULT = 1.7
+        dist_closest_tm = self.fill_radar(radar, self.world.teammates, 0, RADIAL_START, RADIAL_MULT)
+        dist_closest_opp = self.fill_radar(radar, self.world.opponents, 1, RADIAL_START, RADIAL_MULT)
+        self.obs = np.append(radar.flatten(), (dist_closest_tm * 0.5, dist_closest_opp * 0.5, self.output / 40))
+        return self.obs
+
+    def execute(self, action):
+        vec_b_goal = (15.5, 0) - self.world.ball_abs_pos[:2]
+        vec_b_goal_absdir = U.vector_angle(vec_b_goal)
+        rel_direction = action[0] * 60
+        self.output += np.clip(U.normalize_deg(rel_direction - self.output), -45, 45)
+        abs_direction = U.normalize_deg(vec_b_goal_absdir + self.output)
+        return abs_direction

behaviors/custom/Push_RL/Env_LL.py

@@ -0,0 +1,140 @@
+import math
+import numpy as np
+from math_ops.Math_Ops import Math_Ops as U
+from behaviors.custom.Step.Step_Generator import Step_Generator
+from agent.Base_Agent import Base_Agent
+
+
+class Env_LL:
+
+    def __init__(self, base_agent: Base_Agent, step_width=None):
+        self.world = base_agent.world
+        self.obs = np.zeros(78, np.float32)
+        self.STEP_DUR = 8
+        self.STEP_Z_SPAN = 0.02
+        self.STEP_Z_MAX = 0.7
+
+        r = self.world.robot
+        nao_specs = base_agent.inv_kinematics.NAO_SPECS
+        self.leg_length = nao_specs[1] + nao_specs[3]
+        feet_y_dev = nao_specs[0] * step_width
+        sample_time = r.STEPTIME
+        max_ankle_z = nao_specs[5]
+
+        self.step_generator = Step_Generator(feet_y_dev, sample_time, max_ankle_z)
+        self.inv_kinematics = base_agent.inv_kinematics
+        self.DEFAULT_ARMS = np.array([
+            -90, -90, 8, 8, 90, 90, 70, 70], np.float32)
+        self.HL_abs_direction = None
+        self.push_speed = 1
+        self.step_counter = 0
+        self.act = np.zeros(16, np.float32)
+        self.values_l = None
+        self.values_r = None
+
+    def observe(self, init=False):
+        w = self.world
+        r = self.world.robot
+
+        if init:
+            self.step_counter = 0
+            self.act = np.zeros(16, np.float32)
+
+        # 填充观测向量
+        self.obs[0] = min(self.step_counter, 96) / 100
+        self.obs[1] = r.loc_head_z * 3
+        self.obs[2] = r.loc_head_z_vel / 2
+        self.obs[3] = r.imu_torso_roll / 15
+        self.obs[4] = r.imu_torso_pitch / 15
+        self.obs[5:8] = r.gyro / 100
+        self.obs[8:11] = r.acc / 10
+        self.obs[11:17] = r.frp.get('lf', np.zeros(6)) * (10, 10, 10, 0.01, 0.01, 0.01)
+        self.obs[17:23] = r.frp.get('rf', np.zeros(6)) * (10, 10, 10, 0.01, 0.01, 0.01)
+        self.obs[23:43] = r.joints_position[2:22] / 100
+        self.obs[43:63] = r.joints_speed[2:22] / 6.1395
+
+        if init:
+            self.obs[63] = 1
+            self.obs[64] = 1
+            self.obs[65] = 0
+            self.obs[66] = 0
+        else:
+            self.obs[63] = self.step_generator.external_progress
+            self.obs[64] = float(self.step_generator.state_is_left_active)
+            self.obs[65] = float(not self.step_generator.state_is_left_active)
+            self.obs[66] = math.sin((self.step_generator.state_current_ts / self.step_generator.ts_per_step) * math.pi)
+
+        ball_rel_hip_center = self.inv_kinematics.torso_to_hip_transform(w.ball_rel_torso_cart_pos)
+        ball_dist_hip_center = np.linalg.norm(ball_rel_hip_center)
+
+        if init:
+            self.obs[67:70] = (0, 0, 0)
+        elif w.ball_is_visible:
+            self.obs[67:70] = (ball_rel_hip_center - self.obs[70:73]) * 10
+
+        self.obs[70:73] = ball_rel_hip_center
+        self.obs[73] = ball_dist_hip_center * 2
+
+        rel_HL_target = U.normalize_deg(self.HL_abs_direction - r.imu_torso_orientation)
+        self.obs[74] = U.deg_cos(rel_HL_target)
+        self.obs[75] = U.deg_sin(rel_HL_target)
+        self.obs[76] = 2
+        self.obs[77] = 2
+
+        # 找到最近的对手
+        opps_dist = [o.state_horizontal_dist for o in w.opponents]
+        if opps_dist:
+            closest_opp_idx = np.argmin(opps_dist)
+            o = w.opponents[closest_opp_idx]
+
+            if opps_dist[closest_opp_idx] < 1:
+                body_parts_rel_torso_2d_avg = np.zeros(2)
+                weight_sum = 0
+
+                for pos in o.body_parts_cart_rel_pos.values():
+                    bp_rel_torso_2d = r.head_to_body_part_transform('torso', pos)[:2]
+                    weight = math.pow(1e+06, -np.linalg.norm(bp_rel_torso_2d))
+                    body_parts_rel_torso_2d_avg += weight * bp_rel_torso_2d
+                    weight_sum += weight
+
+                if weight_sum > 0:
+                    body_parts_rel_torso_2d_avg /= weight_sum
+                    self.obs[76] = body_parts_rel_torso_2d_avg[0]
+                    self.obs[77] = body_parts_rel_torso_2d_avg[1]
+
+        return self.obs
+
+    def execute_ik(self, l_pos, l_rot, r_pos, r_rot):
+        r = self.world.robot
+        (indices, self.values_l, error_codes) = self.inv_kinematics.leg(
+            l_pos, l_rot, True, dynamic_pose=False)
+        r.set_joints_target_position_direct(indices, self.values_l, harmonize=False)
+
+        (indices, self.values_r, error_codes) = self.inv_kinematics.leg(
+            r_pos, r_rot, False, dynamic_pose=False)
+        r.set_joints_target_position_direct(indices, self.values_r, harmonize=False)
+
+    def execute(self, action):
+        r = self.world.robot
+        self.act = 0.8 * self.act + 0.2 * action * 0.9 * self.push_speed
+
+        (lfy, lfz, rfy, rfz) = self.step_generator.get_target_positions(
+            self.step_counter == 0, self.STEP_DUR, self.STEP_Z_SPAN,
+            self.leg_length * self.STEP_Z_MAX)
+
+        a = self.act
+        l_ankle_pos = (a[0] * 0.025 - 0.01, a[1] * 0.01 + lfy, a[2] * 0.01 + lfz)
+        r_ankle_pos = (a[3] * 0.025 - 0.01, a[4] * 0.01 + rfy, a[5] * 0.01 + rfz)
+
+        l_foot_rot = a[6:9] * (2, 2, 3)
+        r_foot_rot = a[9:12] * (2, 2, 3)
+        l_foot_rot[2] = max(0, l_foot_rot[2] + 18.3)
+        r_foot_rot[2] = min(0, r_foot_rot[2] - 18.3)
+
+        arms = np.copy(self.DEFAULT_ARMS)
+        arms[0:4] += a[12:16] * 4
+
+        self.execute_ik(l_ankle_pos, l_foot_rot, r_ankle_pos, r_foot_rot)
+        r.set_joints_target_position_direct(slice(14, 22), arms, harmonize=False)
+
+        self.step_counter += 1

behaviors/custom/Push_RL/Push_RL.py

@@ -0,0 +1,164 @@
+import pickle
+from behaviors.custom.Push_RL.Env_LL import Env_LL
+from behaviors.custom.Push_RL.Env_HL import Env_HL
+from math_ops.Neural_Network import run_mlp
+from math_ops.Math_Ops import Math_Ops as U
+import numpy as np
+from agent.Base_Agent import Base_Agent
+
+
+class Push_RL:
+
+    def __init__(self, base_agent: Base_Agent):
+        self.world = base_agent.world
+        self.description = 'RL push'
+        self.auto_head = True
+        self.env_LL = Env_LL(base_agent, 0.9 if self.world.robot.type == 3 else 1.2)
+        self.env_HL = Env_HL(base_agent)
+        self.phase = 0
+        self.counter = 0
+        self.behavior = base_agent.behavior
+        self.path_manager = base_agent.path_manager
+        self.inv_kinematics = base_agent.inv_kinematics
+
+        # 模型加载（这部分可能在反编译中丢失）
+        with open(U.get_active_directory([
+                                             "/behaviors/custom/Push_RL/push_LL_R1_X9_49152000_steps.pkl",
+                                             "/behaviors/custom/Push_RL/push_LL_R1_X9_49152000_steps.pkl",
+                                             "/behaviors/custom/Push_RL/push_LL_R1_X9_49152000_steps.pkl",
+                                             "/behaviors/custom/Push_RL/push_LL_R1_X9_49152000_steps.pkl",
+                                             "/behaviors/custom/Push_RL/push_LL_R1_X9_49152000_steps.pkl"
+                                         ][self.world.robot.type]), 'rb') as f:
+            self.model_LL = pickle.load(f)
+        with open(U.get_active_directory([
+                                             "/behaviors/custom/Push_RL/push_HL_R1_X9_1966080_steps.pkl",
+                                             "/behaviors/custom/Push_RL/push_HL_R1_X9_1966080_steps.pkl",
+                                             "/behaviors/custom/Push_RL/push_HL_R1_X9_1966080_steps.pkl",
+                                             "/behaviors/custom/Push_RL/push_HL_R1_X9_1966080_steps.pkl",
+                                             "/behaviors/custom/Push_RL/push_HL_R1_X9_1966080_steps.pkl"
+                                         ][self.world.robot.type]), 'rb') as f:
+            self.model_HL = pickle.load(f)
+    def execute(self, reset, stop=False):
+        ''' Just push the ball autonomously, no target is required '''
+        w = self.world
+        r = self.world.robot
+        bp = w.ball_abs_pos[:2]
+        me = r.loc_head_position[:2]
+        step_gen = self.behavior.get_custom_behavior_object('Walk').env.step_generator
+        reset_push = False
+
+        if reset:
+            self.phase = 0
+            b_rel = w.ball_rel_torso_cart_pos
+            if self.behavior.previous_behavior == 'Dribble':
+                if b_rel[0] < 0.25:
+                    pass
+                else:
+                    self.phase = 0
+            elif abs(b_rel[1]) < 0.07:
+                self.phase = 1
+                reset_push = True
+
+        if self.phase == 0:
+            goal_target = (15.1, np.clip(bp[1], -0.7, 0.7))
+            goal_ori = U.vector_angle(goal_target - bp)
+            vec_me_ball_ori = U.vector_angle(bp - me)
+            rel_curr_angle = U.normalize_deg(vec_me_ball_ori - goal_ori)
+            abs_targ_angle = goal_ori + np.clip(rel_curr_angle, -60, 60)
+
+            if bp[1] > 9:
+                abs_targ_angle = np.clip(abs_targ_angle, -160, -20)
+            elif bp[1] < -9:
+                abs_targ_angle = np.clip(abs_targ_angle, 20, 160)
+
+            if bp[0] > 14:
+                if bp[1] > 1.1:
+                    abs_targ_angle = np.clip(abs_targ_angle, -140, -100)
+                elif bp[1] < -1.1:
+                    abs_targ_angle = np.clip(abs_targ_angle, 100, 140)
+                else:
+                    abs_targ_angle = goal_ori
+
+            ball_dist = np.linalg.norm(bp - me)
+            ori = None if ball_dist > 0.8 else abs_targ_angle
+            (next_pos, next_ori, dist_to_final_target) = self.path_manager.get_path_to_ball(
+                x_ori=abs_targ_angle, x_dev=-0.19, torso_ori=ori)
+
+            b_rel = w.ball_rel_torso_cart_pos
+            ang_diff = abs(U.normalize_deg(abs_targ_angle - r.imu_torso_orientation))
+
+            # 检查是否可以进入阶段1
+            if b_rel[0] >= 0.25 and abs(b_rel[1]) < 0.05 and step_gen.state_is_left_active and step_gen.switch and \
+                    w.time_local_ms - w.ball_abs_pos_last_update < 300 and ang_diff < 10:
+                reset_push = True
+                self.phase += 1
+                self.counter = 0
+            else:
+                dist = max(0.13, dist_to_final_target)
+                reset_walk = reset or self.behavior.previous_behavior != 'Walk'
+                self.behavior.execute_sub_behavior('Walk', reset_walk, next_pos, True,
+                                                next_ori if dist_to_final_target < 1 else None, True, dist)
+
+        if stop:
+            self.phase = 0 # Reset phase on forced stop
+            return True
+
+        if self.phase == 1:
+            # 检查是否要离开场地
+            leaving_field = False
+            if (bp[1] > 9 and r.imu_torso_orientation > 0) or \
+                    (bp[1] < -9 and r.imu_torso_orientation < 0) or \
+                    (bp[0] > 14 and abs(bp[1]) > 1.1):
+                leaving_field = abs(r.imu_torso_orientation) < 90
+
+            ball_hip = self.inv_kinematics.torso_to_hip_transform(w.ball_rel_torso_cart_pos)[:2]
+            dist_ball_our_goal = np.linalg.norm(bp - (-15, 0))
+            dist_us_our_goal = np.linalg.norm(me - (-15, 0))
+
+            # 检查是否丢球
+            lost = abs(ball_hip[1]) > 0.2
+            ball_unseen = w.time_local_ms - w.ball_last_seen >= 400
+            ball_far = np.linalg.norm(ball_hip) > 0.3
+
+            terminal = leaving_field or (dist_ball_our_goal + 0.2 < dist_us_our_goal) or \
+                       (not ball_unseen and ball_far and lost)
+
+            if stop or terminal:
+                self.phase += 1
+            elif self.counter % 25 == 0:
+                obs = self.env_HL.observe(reset_push)
+                action = run_mlp(obs, self.model_HL)
+                self.env_LL.HL_abs_direction = self.env_HL.execute(action)
+
+            self.counter += 1
+            self.env_LL.push_speed = 1
+            obs = self.env_LL.observe(reset_push)
+            action = run_mlp(obs, self.model_LL)
+            self.env_LL.execute(action)
+
+            # 绘制调试信息
+            d = w.draw
+            if d.enabled:
+                vec = U.vector_from_angle(self.env_LL.HL_abs_direction)
+                d.line(me, me + vec, 4, d.Color.red, 'opp_vec')
+
+            return False
+
+        if self.phase > 1:
+            WIND_DOWN_STEPS = 50
+            self.env_LL.push_speed = 1 - self.phase / WIND_DOWN_STEPS
+            self.env_LL.HL_abs_direction = r.imu_torso_orientation
+            obs = self.env_LL.observe(reset_push)
+            action = run_mlp(obs, self.model_LL)
+            self.env_LL.execute(action)
+            self.phase += 1
+
+            if self.phase >= WIND_DOWN_STEPS - 5 or np.linalg.norm(r.get_head_abs_vel(4)) < 0.15:
+                self.phase = 0
+                return True
+
+        return False
+
+    def is_ready(self):
+        ''' Returns True if Push Behavior is ready to start under current game/robot conditions '''
+        return True

logs/2024-09-06_19.02.40__NIEEQB/FCPortugal_1.log

@@ -0,0 +1,2 @@
+{Fri 19:02:40 Step:2513} World_Parser.py: Received field line with NaNs [ 2.98 43.85 21.32   nan   nan   nan]
+{Fri 19:03:14 Step:4213} World_Parser.py: Received field line with NaNs [ 3.38 35.5  20.11   nan   nan   nan]

scripts/gyms/dribble.py

@@ -0,0 +1,396 @@
+import math
+import random
+
+from agent.Base_Agent import Base_Agent as Agent
+from behaviors.custom.Step.Step import Step
+from world.commons.Draw import Draw
+from stable_baselines3 import PPO
+from stable_baselines3.common.vec_env import SubprocVecEnv
+from scripts.commons.Server import Server
+from scripts.commons.Train_Base import Train_Base
+from time import sleep
+import os, gym
+import numpy as np
+from math_ops.Math_Ops import Math_Ops as U
+from math_ops.Math_Ops import Math_Ops as M
+from behaviors.custom.Step.Step_Generator import Step_Generator
+
+'''
+Objective:
+Learn how to run forward using step primitive
+----------
+- class Basic_Run: implements an OpenAI custom gym
+- class Train:  implements algorithms to train a new model or test an existing model
+'''
+
+
+class dribble(gym.Env):
+    def __init__(self, ip, server_p, monitor_p, r_type, enable_draw) -> None:
+        self.abs_ori = 75
+        self.ball_dist_hip_center_2d = 0
+        self.ball_dist_hip_center = None
+        self.internal_rel_orientation = None
+        self.dribble_speed = 1
+        self.gym_last_internal_abs_ori = None
+        self.internal_target_vel = None
+        self.Gen_player_pos = None
+        self.internal_target = None
+        self.values_l = None
+        self.values_r = None
+        self.reset_time = None
+        self.behavior = None
+        self.robot_type = r_type
+        self.kick_ori = 0
+        self.terminal = False
+        # Args: Server IP, Agent Port, Monitor Port, Uniform No., Robot Type, Team Name, Enable Log, Enable Draw
+        self.player = Agent(ip, server_p, monitor_p, 1, self.robot_type, "Gym", True, True)
+        
+        self.step_counter = 0  # to limit episode size
+        self.ik = self.player.inv_kinematics
+        self.dribble_rel_orientation = 0  # relative to imu_torso_orientation (in degrees)
+        # Step behavior defaults
+        self.STEP_DUR = 8
+        self.STEP_Z_SPAN = 0.02
+        self.STEP_Z_MAX = 0.70
+        nao_specs = self.ik.NAO_SPECS
+        self.leg_length = nao_specs[1] + nao_specs[3]  # upper leg height + lower leg height
+        feet_y_dev = nao_specs[0] * 1.12  # wider step
+        sample_time = self.player.world.robot.STEPTIME
+        max_ankle_z = nao_specs[5]
+        self.step_generator = Step_Generator(feet_y_dev, sample_time, max_ankle_z)
+        self.DEFAULT_ARMS = np.array([-90, -90, 8, 8, 90, 90, 70, 70], np.float32)
+
+        self.act = np.zeros(16, np.float32)  # memory variable
+        self.path_manager = self.player.path_manager
+
+        # State space
+        obs_size = 76
+        self.obs = np.zeros(obs_size, np.float32)
+        self.observation_space = gym.spaces.Box(low=np.full(obs_size, -np.inf, np.float32),
+                                                high=np.full(obs_size, np.inf, np.float32), dtype=np.float32)
+
+        self.ball_x_center = 0.21
+        self.ball_y_center = -0.045
+        # Action space
+
+        MAX = np.finfo(np.float32).max
+        self.no_of_actions = act_size = 16
+        self.action_space = gym.spaces.Box(low=np.full(act_size, -MAX, np.float32),
+                                           high=np.full(act_size, MAX, np.float32), dtype=np.float32)
+
+        # Place ball far away to keep landmarks in FoV (head follows ball while using Step behavior)
+        self.player.scom.unofficial_move_ball((14, 0, 0.042))
+
+        self.player.scom.unofficial_set_play_mode("PlayOn")
+        self.player.scom.unofficial_move_ball((0, 0, 0))
+
+    def observe(self, init=False, virtual_ball=False):
+        r = self.player.world.robot
+        w = self.player.world
+
+        if init:  # reset variables
+            self.step_counter = 0
+            self.act = np.zeros(16, np.float32)  # memory variable
+
+        # index       observation              naive normalization
+        self.obs[0] = min(self.step_counter, 12 * 8) / 100  # simple counter: 0,1,2,3...
+        self.obs[1] = r.loc_head_z * 3  # z coordinate (torso)
+        self.obs[2] = r.loc_head_z_vel / 2  # z velocity (torso)
+        self.obs[3] = r.imu_torso_roll / 15  # absolute torso roll  in deg
+        self.obs[4] = r.imu_torso_pitch / 15  # absolute torso pitch in deg
+        self.obs[5:8] = r.gyro / 100  # gyroscope
+        self.obs[8:11] = r.acc / 10  # accelerometer
+
+        self.obs[11:17] = r.frp.get('lf', np.zeros(6)) * (10, 10, 10, 0.01, 0.01,
+                                                          0.01)  # left foot: relative point of origin (p) and force vector (f) -> (px,py,pz,fx,fy,fz)*
+        self.obs[17:23] = r.frp.get('rf', np.zeros(6)) * (10, 10, 10, 0.01, 0.01,
+                                                          0.01)  # right foot: relative point of origin (p) and force vector (f) -> (px,py,pz,fx,fy,fz)*
+        # *if foot is not touching the ground, then (px=0,py=0,pz=0,fx=0,fy=0,fz=0)
+
+        self.obs[23:43] = r.joints_position[2:22] / 100  # position of all joints except head & toes (for robot type 4)
+        self.obs[43:63] = r.joints_speed[2:22] / 6.1395  # speed of    all joints except head & toes (for robot type 4)
+
+        '''
+        Expected observations for walking state:
+        Time step        R  0   1   2   3   4   5   6   7   0
+        Progress         1  0 .14 .28 .43 .57 .71 .86   1   0
+        Left leg active  T  F   F   F   F   F   F   F   F   T
+        '''
+
+        if init:  # the walking parameters refer to the last parameters in effect (after a reset, they are pointless)
+            self.obs[63] = 1  # step progress
+            self.obs[64] = 1  # 1 if left  leg is active
+            self.obs[65] = 0  # 1 if right leg is active
+            self.obs[66] = 0
+        else:
+            self.obs[63] = self.step_generator.external_progress  # step progress
+            self.obs[64] = float(self.step_generator.state_is_left_active)  # 1 if left  leg is active
+            self.obs[65] = float(not self.step_generator.state_is_left_active)  # 1 if right leg is active
+            self.obs[66] = math.sin(self.step_generator.state_current_ts / self.step_generator.ts_per_step * math.pi)
+        #ball
+        ball_rel_hip_center = self.ik.torso_to_hip_transform(w.ball_rel_torso_cart_pos)
+        ball_dist_hip_center = np.linalg.norm(ball_rel_hip_center)
+        if init:
+            self.obs[67:70] = (0, 0, 0)  # Initial velocity is 0
+        elif w.ball_is_visible:
+            self.obs[67:70] = (ball_rel_hip_center - self.obs[
+                                                     70:73]) * 10  # Ball velocity, relative to ankle's midpoint
+
+        self.obs[70:73] = ball_rel_hip_center  # Ball position, relative to hip
+        self.obs[73] = self.ball_dist_hip_center * 2
+
+        if virtual_ball:  # simulate the ball between the robot's feet
+            self.obs[67:74] = (0, 0, 0, 0.05, 0, -0.175, 0.36)
+
+        '''
+        Create internal target with a smoother variation
+        '''
+
+        MAX_ROTATION_DIFF = 20  # max difference (degrees) per visual step
+        MAX_ROTATION_DIST = 80
+
+        if init:
+            self.internal_rel_orientation = 0
+            self.internal_target_vel = 0
+            self.gym_last_internal_abs_ori = r.imu_torso_orientation  # for training purposes (reward)
+
+        # ---------------------------------------------------------------- compute internal target
+
+        if w.vision_is_up_to_date:
+            previous_internal_rel_orientation = np.copy(self.internal_rel_orientation)
+
+            internal_ori_diff = np.clip(M.normalize_deg(self.dribble_rel_orientation - self.internal_rel_orientation),
+                                        -MAX_ROTATION_DIFF, MAX_ROTATION_DIFF)
+            self.internal_rel_orientation = np.clip(M.normalize_deg(self.internal_rel_orientation + internal_ori_diff),
+                                                    -MAX_ROTATION_DIST, MAX_ROTATION_DIST)
+
+            # Observations
+            self.internal_target_vel = self.internal_rel_orientation - previous_internal_rel_orientation
+
+            self.gym_last_internal_abs_ori = self.internal_rel_orientation + r.imu_torso_orientation
+        # ----------------------------------------------------------------- observations
+
+        self.obs[74] = self.internal_rel_orientation / MAX_ROTATION_DIST
+        self.obs[75] = self.internal_target_vel / MAX_ROTATION_DIFF
+
+        return self.obs
+
+    def execute_ik(self, l_pos, l_rot, r_pos, r_rot):
+        r = self.player.world.robot
+        # Apply IK to each leg + Set joint targets
+
+        # Left leg
+        indices, self.values_l, error_codes = self.ik.leg(l_pos, l_rot, True, dynamic_pose=False)
+
+        r.set_joints_target_position_direct(indices, self.values_l, harmonize=False)
+
+        # Right leg
+        indices, self.values_r, error_codes = self.ik.leg(r_pos, r_rot, False, dynamic_pose=False)
+
+        r.set_joints_target_position_direct(indices, self.values_r, harmonize=False)
+
+    def sync(self):
+        ''' Run a single simulation step '''
+        r = self.player.world.robot
+        self.player.scom.commit_and_send(r.get_command())
+        self.player.scom.receive()
+
+    def reset(self):
+        # print("reset")
+        '''
+        Reset and stabilize the robot
+        Note: for some behaviors it would be better to reduce stabilization or add noise
+        '''
+        self.abs_ori = 45
+        self.player.scom.unofficial_set_play_mode("PlayOn")
+
+        Gen_ball_pos = [- 9, 0, 0]
+        self.Gen_player_pos = (Gen_ball_pos[0] - 1.5, Gen_ball_pos[1], 0.5)
+        self.player.scom.unofficial_move_ball((Gen_ball_pos[0], Gen_ball_pos[1], Gen_ball_pos[2]))
+        self.step_counter = 0
+        self.behavior = self.player.behavior
+        r = self.player.world.robot
+        w = self.player.world
+        t = w.time_local_ms
+        self.reset_time = t
+        self.b_rel = w.ball_rel_torso_cart_pos[:2]
+        self.path_manager = self.player.path_manager
+
+        for _ in range(25):
+            self.player.scom.unofficial_beam(self.Gen_player_pos, 0)  # beam player continuously (floating above ground)
+            self.player.behavior.execute("Zero_Bent_Knees")
+            self.sync()
+
+        # beam player to ground
+        self.player.scom.unofficial_beam(self.Gen_player_pos, 0)
+        r.joints_target_speed[
+            0] = 0.01  # move head to trigger physics update (rcssserver3d bug when no joint is moving)
+        self.sync()
+
+        # stabilize on ground
+        for _ in range(7):
+            self.player.behavior.execute("Zero_Bent_Knees")
+            self.sync()
+        # walk to ball
+        while True:
+            self.b_rel = w.ball_rel_torso_cart_pos[:2]
+            if self.player.behavior.is_ready("Get_Up"):
+                self.player.behavior.execute_to_completion("Get_Up")
+            if 0.26 > self.b_rel[0] > 0.18 and abs(self.b_rel[1]) < 0.04 and w.ball_is_visible:
+                break
+            else:
+                if self.player.behavior.is_ready("Get_Up"):
+                    self.player.behavior.execute_to_completion("Get_Up")
+
+                reset_walk = self.behavior.previous_behavior != "Walk"  # reset walk if it wasn't the previous behavior
+                rel_target = self.b_rel + (-0.23, 0)  # relative target is a circle (center: ball, radius:0.23m)
+                rel_ori = M.vector_angle(self.b_rel)  # relative orientation of ball, NOT the target!
+                dist = max(0.08, np.linalg.norm(rel_target) * 0.7)  # slow approach
+                self.behavior.execute_sub_behavior("Walk", reset_walk, rel_target, False, rel_ori, False,
+                                                   dist)  # target, is_target_abs, ori, is_ori_abs, distance
+
+                self.sync()
+
+        self.act = np.zeros(self.no_of_actions, np.float32)
+        return self.observe(True)
+
+    def render(self, mode='human', close=False):
+        return
+
+    def close(self):
+        Draw.clear_all()
+        self.player.terminate()
+
+    def step(self, action):
+
+        r = (self.
+             player.world.robot)
+
+        w = self.player.world
+        d = w.draw
+        if w.ball_abs_pos[1] > 0:  #
+            dribble_target = (15, 5)
+        else:
+            dribble_target = (15, -5)
+
+        self.dribble_rel_orientation = self.path_manager.get_dribble_path(optional_2d_target=dribble_target)[1]
+        # exponential moving average
+        self.act = 0.85 * self.act + 0.15 * action * 0.7 * 0.95 * self.dribble_speed
+
+        # execute Step behavior to extract the target positions of each leg (we will override these targets)
+        lfy, lfz, rfy, rfz = self.step_generator.get_target_positions(self.step_counter == 0, self.STEP_DUR,
+                                                                      self.STEP_Z_SPAN,
+                                                                      self.leg_length * self.STEP_Z_MAX)
+
+        # Leg IK
+        a = self.act
+        l_ankle_pos = (a[0] * 0.025 - 0.01, a[1] * 0.01 + lfy, a[2] * 0.01 + lfz)
+        r_ankle_pos = (a[3] * 0.025 - 0.01, a[4] * 0.01 + rfy, a[5] * 0.01 + rfz)
+        l_foot_rot = a[6:9] * (2, 2, 3)
+        r_foot_rot = a[9:12] * (2, 2, 3)
+
+        # Limit leg yaw/pitch (and add bias)
+        l_foot_rot[2] = max(0, l_foot_rot[2] + 18.3)
+        r_foot_rot[2] = min(0, r_foot_rot[2] - 18.3)
+
+        # Arms actions
+        arms = np.copy(self.DEFAULT_ARMS)  # default arms pose
+        arms[0:4] += a[12:16] * 4  # arms pitch+roll
+
+        # Set target positions
+        self.execute_ik(l_ankle_pos, l_foot_rot, r_ankle_pos, r_foot_rot)  # legs
+        r.set_joints_target_position_direct(slice(14, 22), arms, harmonize=False)  # arms
+        self.sync()
+        self.step_counter += 1
+
+        # 收集观测
+        obs = self.observe()
+        angle_rad = np.radians(self.gym_last_internal_abs_ori)  # 将角度转换为弧度
+        unit_vector = np.array([np.cos(angle_rad), np.sin(angle_rad)])  # 计算单位向量
+        if np.linalg.norm(w.ball_cheat_abs_vel[:2]) != 0:
+            cos_theta = np.dot(unit_vector, w.ball_cheat_abs_vel[:2]) / (
+                    np.linalg.norm(unit_vector) * np.linalg.norm(w.ball_cheat_abs_vel[:2]))
+        else:
+            cos_theta = 0
+        # 计算奖励
+        reward = np.linalg.norm(w.ball_cheat_abs_vel) * cos_theta
+
+        if self.ball_dist_hip_center_2d < 0.115:
+            reward = 0
+        # 判断终止
+        if self.player.behavior.is_ready("Get_Up"):
+            self.terminal = True
+        elif w.time_local_ms - self.reset_time > 7500 * 3 or np.linalg.norm(
+                w.ball_cheat_abs_pos[:2] - r.loc_head_position[:2]) > 0.45 or not w.is_ball_abs_pos_from_vision:
+            self.terminal = True
+        else:
+            self.terminal = False
+        return obs, reward, self.terminal, {}
+
+
+class Train(Train_Base):
+    def __init__(self, script) -> None:
+        super().__init__(script)
+
+    def train(self, args):
+
+        # --------------------------------------- Learning parameters
+        n_envs = min(1, os.cpu_count())
+        n_steps_per_env = 1024  # RolloutBuffer is of size (n_steps_per_env * n_envs)
+        minibatch_size = 64  # should be a factor of (n_steps_per_env * n_envs)
+        total_steps = 50000000
+        learning_rate = 3e-4
+        folder_name = f'Sprint_R{self.robot_type}'
+        model_path = f'./scripts/gyms/logs/{folder_name}/'
+
+        # print("Model path:", model_path)
+
+        # --------------------------------------- Run algorithm
+        def init_env(i_env):
+            def thunk():
+                return dribble(self.ip, self.server_p + i_env, self.monitor_p_1000 + i_env, self.robot_type, False)
+
+            return thunk
+
+        servers = Server(self.server_p, self.monitor_p_1000, n_envs + 1)  # include 1 extra server for testing
+
+        env = SubprocVecEnv([init_env(i) for i in range(n_envs)])
+        eval_env = SubprocVecEnv([init_env(n_envs)])
+
+        try:
+            if "model_file" in args:  # retrain
+                model = PPO.load(args["model_file"], env=env, device="cuda", n_envs=n_envs, n_steps=n_steps_per_env,
+                                 batch_size=minibatch_size, learning_rate=learning_rate)
+            else:  # train new model
+                model = PPO("MlpPolicy", env=env, verbose=1, n_steps=n_steps_per_env, batch_size=minibatch_size,
+                            learning_rate=learning_rate, device="cuda")
+
+            model_path = self.learn_model(model, total_steps, model_path, eval_env=eval_env,
+                                          eval_freq=n_steps_per_env * 20, save_freq=n_steps_per_env * 20,
+                                          backup_env_file=__file__)
+        except KeyboardInterrupt:
+            sleep(1)  # wait for child processes
+            ("\nctrl+c pressed, aborting...\n")
+            servers.kill()
+            return
+
+        env.close()
+        eval_env.close()
+        servers.kill()
+
+    def test(self, args):
+
+        # Uses different server and monitor ports
+        server = Server(self.server_p - 1, self.monitor_p, 1)
+        env = dribble(self.ip, self.server_p - 1, self.monitor_p, self.robot_type, True)
+        model = PPO.load(args["model_file"], env=env)
+
+        try:
+            self.export_model(args["model_file"], args["model_file"] + ".pkl",
+                              False)  # Export to pkl to create custom behavior
+            self.test_model(model, env, log_path=args["folder_dir"], model_path=args["folder_dir"])
+        except KeyboardInterrupt:
+            print()
+
+        env.close()
+        server.kill()

scripts/utils/gail/gail.py

@@ -0,0 +1,403 @@
+import math
+import random
+import pickle
+from agent.Base_Agent import Base_Agent as Agent
+from behaviors.custom.Step.Step import Step
+from world.commons.Draw import Draw
+from stable_baselines3 import PPO
+from stable_baselines3.common.vec_env import SubprocVecEnv
+from scripts.commons.Server import Server
+from scripts.commons.Train_Base import Train_Base
+from time import sleep
+import os, gym
+import numpy as np
+from math_ops.Math_Ops import Math_Ops as U
+from math_ops.Math_Ops import Math_Ops as M
+from behaviors.custom.Step.Step_Generator import Step_Generator
+from sb3_contrib import GAIL
+
+'''
+Objective:
+Learn how to run forward using step primitive
+----------
+- class Basic_Run: implements an OpenAI custom gym
+- class Train:  implements algorithms to train a new model or test an existing model
+'''
+
+
+class dribble(gym.Env):
+    def __init__(self, ip, server_p, monitor_p, r_type, enable_draw) -> None:
+        self.expert_data = dict()
+        self.abs_ori = 75
+        self.ball_dist_hip_center_2d = 0
+        self.ball_dist_hip_center = None
+        self.internal_rel_orientation = None
+        self.dribble_speed = 1
+        self.gym_last_internal_abs_ori = None
+        self.internal_target_vel = None
+        self.Gen_player_pos = None
+        self.internal_target = None
+        self.values_l = None
+        self.values_r = None
+        self.reset_time = None
+        self.behavior = None
+        self.robot_type = r_type
+        self.kick_ori = 0
+        self.terminal = False
+        # Args: Server IP, Agent Port, Monitor Port, Uniform No., Robot Type, Team Name, Enable Log, Enable Draw
+        self.player = Agent(ip, server_p, monitor_p, 1, self.robot_type, "Gym", True, True)
+
+        self.step_counter = 0  # to limit episode size
+        self.ik = self.player.inv_kinematics
+        self.dribble_rel_orientation = 0  # relative to imu_torso_orientation (in degrees)
+        # Step behavior defaults
+        self.STEP_DUR = 8
+        self.STEP_Z_SPAN = 0.02
+        self.STEP_Z_MAX = 0.70
+        nao_specs = self.ik.NAO_SPECS
+        self.leg_length = nao_specs[1] + nao_specs[3]  # upper leg height + lower leg height
+        feet_y_dev = nao_specs[0] * 1.12  # wider step
+        sample_time = self.player.world.robot.STEPTIME
+        max_ankle_z = nao_specs[5]
+        self.step_generator = Step_Generator(feet_y_dev, sample_time, max_ankle_z)
+        self.DEFAULT_ARMS = np.array([-90, -90, 8, 8, 90, 90, 70, 70], np.float32)
+
+        self.act = np.zeros(16, np.float32)  # memory variable
+        self.path_manager = self.player.path_manager
+
+        # State space
+        obs_size = 76
+        self.obs = np.zeros(obs_size, np.float32)
+        self.observation_space = gym.spaces.Box(low=np.full(obs_size, -np.inf, np.float32),
+                                                high=np.full(obs_size, np.inf, np.float32), dtype=np.float32)
+
+        self.ball_x_center = 0.21
+        self.ball_y_center = -0.045
+        # Action space
+
+        MAX = np.finfo(np.float32).max
+        self.no_of_actions = act_size = 16
+        self.action_space = gym.spaces.Box(low=np.full(act_size, -MAX, np.float32),
+                                           high=np.full(act_size, MAX, np.float32), dtype=np.float32)
+
+        # Place ball far away to keep landmarks in FoV (head follows ball while using Step behavior)
+        self.player.scom.unofficial_move_ball((14, 0, 0.042))
+
+        self.player.scom.unofficial_set_play_mode("PlayOn")
+        self.player.scom.unofficial_move_ball((0, 0, 0))
+
+    def observe(self, init=False, virtual_ball=False):
+        r = self.player.world.robot
+        w = self.player.world
+
+        if init:  # reset variables
+            self.step_counter = 0
+            self.act = np.zeros(16, np.float32)  # memory variable
+
+        # index       observation              naive normalization
+        self.obs[0] = min(self.step_counter, 12 * 8) / 100  # simple counter: 0,1,2,3...
+        self.obs[1] = r.loc_head_z * 3  # z coordinate (torso)
+        self.obs[2] = r.loc_head_z_vel / 2  # z velocity (torso)
+        self.obs[3] = r.imu_torso_roll / 15  # absolute torso roll  in deg
+        self.obs[4] = r.imu_torso_pitch / 15  # absolute torso pitch in deg
+        self.obs[5:8] = r.gyro / 100  # gyroscope
+        self.obs[8:11] = r.acc / 10  # accelerometer
+
+        self.obs[11:17] = r.frp.get('lf', np.zeros(6)) * (10, 10, 10, 0.01, 0.01,
+                                                          0.01)  # left foot: relative point of origin (p) and force vector (f) -> (px,py,pz,fx,fy,fz)*
+        self.obs[17:23] = r.frp.get('rf', np.zeros(6)) * (10, 10, 10, 0.01, 0.01,
+                                                          0.01)  # right foot: relative point of origin (p) and force vector (f) -> (px,py,pz,fx,fy,fz)*
+        # *if foot is not touching the ground, then (px=0,py=0,pz=0,fx=0,fy=0,fz=0)
+
+        self.obs[23:43] = r.joints_position[2:22] / 100  # position of all joints except head & toes (for robot type 4)
+        self.obs[43:63] = r.joints_speed[2:22] / 6.1395  # speed of    all joints except head & toes (for robot type 4)
+
+        if init:  # the walking parameters refer to the last parameters in effect (after a reset, they are pointless)
+            self.obs[63] = 1  # step progress
+            self.obs[64] = 1  # 1 if left  leg is active
+            self.obs[65] = 0  # 1 if right leg is active
+            self.obs[66] = 0
+        else:
+            self.obs[63] = self.step_generator.external_progress  # step progress
+            self.obs[64] = float(self.step_generator.state_is_left_active)  # 1 if left  leg is active
+            self.obs[65] = float(not self.step_generator.state_is_left_active)  # 1 if right leg is active
+            self.obs[66] = math.sin(self.step_generator.state_current_ts / self.step_generator.ts_per_step * math.pi)
+
+        # Ball
+        ball_rel_hip_center = self.ik.torso_to_hip_transform(w.ball_rel_torso_cart_pos)
+        self.ball_dist_hip_center = np.linalg.norm(ball_rel_hip_center)
+        ball_rel_torso_xy = w.ball_rel_torso_cart_pos[:2]  # 取X和Y分量
+        self.ball_dist_hip_center_2d = np.linalg.norm(ball_rel_torso_xy)
+
+        if init:
+            self.obs[67:70] = (0, 0, 0)  # Initial velocity is 0
+        elif w.ball_is_visible:
+            self.obs[67:70] = (ball_rel_hip_center - self.obs[
+                                                     70:73]) * 10  # Ball velocity, relative to ankle's midpoint
+
+        self.obs[70:73] = ball_rel_hip_center  # Ball position, relative to hip
+        self.obs[73] = self.ball_dist_hip_center * 2
+
+        if virtual_ball:  # simulate the ball between the robot's feet
+            self.obs[67:74] = (0, 0, 0, 0.05, 0, -0.175, 0.36)
+
+        '''
+        Create internal target with a smoother variation
+        '''
+
+        MAX_ROTATION_DIFF = 20  # max difference (degrees) per visual step
+        MAX_ROTATION_DIST = 80
+
+        if init:
+            self.internal_rel_orientation = 0
+            self.internal_target_vel = 0
+            self.gym_last_internal_abs_ori = r.imu_torso_orientation  # for training purposes (reward)
+
+        # ---------------------------------------------------------------- compute internal target
+
+        if w.vision_is_up_to_date:
+            previous_internal_rel_orientation = np.copy(self.internal_rel_orientation)
+
+            internal_ori_diff = np.clip(M.normalize_deg(self.dribble_rel_orientation - self.internal_rel_orientation),
+                                        -MAX_ROTATION_DIFF, MAX_ROTATION_DIFF)
+            self.internal_rel_orientation = np.clip(M.normalize_deg(self.internal_rel_orientation + internal_ori_diff),
+                                                    -MAX_ROTATION_DIST, MAX_ROTATION_DIST)
+
+            # Observations
+            self.internal_target_vel = self.internal_rel_orientation - previous_internal_rel_orientation
+
+            self.gym_last_internal_abs_ori = self.internal_rel_orientation + r.imu_torso_orientation
+        # ----------------------------------------------------------------- observations
+
+        self.obs[74] = self.internal_rel_orientation / MAX_ROTATION_DIST
+        self.obs[75] = self.internal_target_vel / MAX_ROTATION_DIFF
+
+        return self.obs
+
+    def sync(self):
+        ''' Run a single simulation step '''
+        r = self.player.world.robot
+        self.player.scom.commit_and_send(r.get_command())
+        self.player.scom.receive()
+
+    def reset(self):
+        # print("reset")
+        '''
+        Reset and stabilize the robot
+        Note: for some behaviors it would be better to reduce stabilization or add noise
+        '''
+        self.abs_ori = 45
+        self.player.scom.unofficial_set_play_mode("PlayOn")
+
+        Gen_ball_pos = [- 9, 0, 0]
+        self.Gen_player_pos = (Gen_ball_pos[0] - 1.5, Gen_ball_pos[1], 0.5)
+        self.player.scom.unofficial_move_ball((Gen_ball_pos[0], Gen_ball_pos[1], Gen_ball_pos[2]))
+        self.step_counter = 0
+        self.behavior = self.player.behavior
+        r = self.player.world.robot
+        w = self.player.world
+        t = w.time_local_ms
+        self.reset_time = t
+        self.b_rel = w.ball_rel_torso_cart_pos[:2]
+        self.path_manager = self.player.path_manager
+
+        for _ in range(25):
+            self.player.scom.unofficial_beam(self.Gen_player_pos, 0)  # beam player continuously (floating above ground)
+            self.player.behavior.execute("Zero_Bent_Knees")
+            self.sync()
+
+        # beam player to ground
+        self.player.scom.unofficial_beam(self.Gen_player_pos, 0)
+        r.joints_target_speed[
+            0] = 0.01  # move head to trigger physics update (rcssserver3d bug when no joint is moving)
+        self.sync()
+
+        # stabilize on ground
+        for _ in range(7):
+            self.player.behavior.execute("Zero_Bent_Knees")
+            self.sync()
+        # walk to ball
+        while True:
+            self.b_rel = w.ball_rel_torso_cart_pos[:2]
+            if self.player.behavior.is_ready("Get_Up"):
+                self.player.behavior.execute_to_completion("Get_Up")
+            if 0.26 > self.b_rel[0] > 0.18 and abs(self.b_rel[1]) < 0.04 and w.ball_is_visible:
+                break
+            else:
+                if self.player.behavior.is_ready("Get_Up"):
+                    self.player.behavior.execute_to_completion("Get_Up")
+
+                reset_walk = self.behavior.previous_behavior != "Walk"  # reset walk if it wasn't the previous behavior
+                rel_target = self.b_rel + (-0.23, 0)  # relative target is a circle (center: ball, radius:0.23m)
+                rel_ori = M.vector_angle(self.b_rel)  # relative orientation of ball, NOT the target!
+                dist = max(0.08, np.linalg.norm(rel_target) * 0.7)  # slow approach
+                self.behavior.execute_sub_behavior("Walk", reset_walk, rel_target, False, rel_ori, False,
+                                                   dist)  # target, is_target_abs, ori, is_ori_abs, distance
+
+                self.sync()
+
+        self.act = np.zeros(self.no_of_actions, np.float32)
+        return self.observe(True)
+
+    def render(self, mode='human', close=False):
+        return
+
+    def close(self):
+        Draw.clear_all()
+        self.player.terminate()
+
+    def run_mlp(self, obs, weights, activation_function="tanh"):
+        '''
+        Run multilayer perceptron using numpy
+
+        Parameters
+        ----------
+        obs : ndarray
+            float32 array with neural network inputs
+        weights : list
+            list of MLP layers of type (bias, kernel)
+        activation_function : str
+            activation function for hidden layers
+            set to "none" to disable
+        '''
+
+        obs = obs.astype(np.float32, copy=False)
+        out = obs
+
+        for w in weights[:-1]:  # for each hidden layer
+            out = np.matmul(w[1], out) + w[0]
+            if activation_function == "tanh":
+                np.tanh(out, out=out)
+            elif activation_function != "none":
+                raise NotImplementedError
+        return np.matmul(weights[-1][1], out) + weights[-1][0]  # final layer
+
+    def step(self, action):
+
+        r = (self.
+             player.world.robot)
+
+        w = self.player.world
+        d = w.draw
+        if w.ball_abs_pos[1] > 0:  #
+            dribble_target = (15, 5)
+        else:
+            dribble_target = (15, -5)
+        with open(M.get_active_directory([
+                                             "/home/her-darling/APOLLO_PRO/TEST_Dribble/scripts/utils/gail/dribble_long_R1_00_178M.pkl",
+                                             "/home/her-darling/APOLLO_PRO/TEST_Dribble/scripts/utils/gail/dribble_long_R1_00_178M.pkl",
+                                             "/home/her-darling/APOLLO_PRO/TEST_Dribble/scripts/utils/gail/dribble_long_R1_00_178M.pkl",
+                                             "/home/her-darling/APOLLO_PRO/TEST_Dribble/scripts/utils/gail/dribble_long_R1_00_178M.pkl",
+                                             "/home/her-darling/APOLLO_PRO/TEST_Dribble/scripts/utils/gail/dribble_long_R1_00_178M.pkl"
+                                         ][r.type]), 'rb') as f:
+            model = pickle.load(f)
+        # 收集观测
+        obs = self.observe()
+        action = self.run_mlp(obs, model)
+        self.sync()
+        self.step_counter += 1
+        angle_rad = np.radians(self.gym_last_internal_abs_ori)  # 将角度转换为弧度
+        unit_vector = np.array([np.cos(angle_rad), np.sin(angle_rad)])  # 计算单位向量
+        if np.linalg.norm(w.ball_cheat_abs_vel[:2]) != 0:
+            cos_theta = np.dot(unit_vector, w.ball_cheat_abs_vel[:2]) / (
+                    np.linalg.norm(unit_vector) * np.linalg.norm(w.ball_cheat_abs_vel[:2]))
+        else:
+            cos_theta = 0
+        #获取专家数据
+        self.expert_data = {
+            "actions": action,
+            "obs": obs
+        }
+        # 计算奖励
+        reward = np.linalg.norm(w.ball_cheat_abs_vel) * cos_theta
+
+        if self.ball_dist_hip_center_2d < 0.115:
+            reward = 0
+        # 判断终止
+        if self.player.behavior.is_ready("Get_Up"):
+            self.terminal = True
+        elif w.time_local_ms - self.reset_time > 7500 * 3 or np.linalg.norm(
+                w.ball_cheat_abs_pos[:2] - r.loc_head_position[:2]) > 0.45 or not w.is_ball_abs_pos_from_vision:
+            self.terminal = True
+        else:
+            self.terminal = False
+        return obs, reward, self.terminal, {}
+
+
+class Train(Train_Base):
+    def __init__(self, script) -> None:
+        super().__init__(script)
+
+    def train(self, args):
+
+        # --------------------------------------- Learning parameters
+        n_envs = min(1, os.cpu_count())
+        n_steps_per_env = 1024  # RolloutBuffer is of size (n_steps_per_env * n_envs)
+        minibatch_size = 64  # should be a factor of (n_steps_per_env * n_envs)
+        total_steps = 50000000
+        learning_rate = 3e-4
+        folder_name = f'Sprint_R{self.robot_type}'
+        model_path = f'./scripts/gyms/logs/{folder_name}/'
+        expert_data = dribble.__init__().expert_data
+
+        # print("Model path:", model_path)
+
+        # --------------------------------------- Run algorithm
+        def init_env(i_env):
+            def thunk():
+                return dribble(self.ip, self.server_p + i_env, self.monitor_p_1000 + i_env, self.robot_type, False)
+
+            return thunk
+
+        servers = Server(self.server_p, self.monitor_p_1000, n_envs + 1)  # include 1 extra server for testing
+
+        env = SubprocVecEnv([init_env(i) for i in range(n_envs)])
+        eval_env = SubprocVecEnv([init_env(n_envs)])
+
+        try:
+            if "model_file" in args:  # retrain
+                model = GAIL.load(args["model_file"], env=env, device="cuda", n_envs=n_envs, n_steps=n_steps_per_env,
+                                 batch_size=minibatch_size, learning_rate=learning_rate)
+            else:  # train new model
+                model = GAIL("MlpPolicy", env=env, expert_dataset=expert_data, verbose=1, n_steps=n_steps_per_env, batch_size=minibatch_size,
+                            learning_rate=learning_rate, device="cuda")
+
+            model_path = self.learn_model(model, total_steps, model_path, eval_env=eval_env,
+                                          eval_freq=n_steps_per_env * 20, save_freq=n_steps_per_env * 20,
+                                          backup_env_file=__file__)
+
+# model = GAIL(
+#     "MlpPolicy",
+#     env,
+#     expert_dataset=expert_data,  # 提供专家数据
+#     verbose=1,
+#     n_steps=2048,  # 每次更新的步数
+#     batch_size=64,  # 批量大小
+#     learning_rate=3e-4
+# )
+        except KeyboardInterrupt:
+            sleep(1)  # wait for child processes
+            ("\nctrl+c pressed, aborting...\n")
+            servers.kill()
+            return
+
+        env.close()
+        eval_env.close()
+        servers.kill()
+
+    def test(self, args):
+
+        # Uses different server and monitor ports
+        server = Server(self.server_p - 1, self.monitor_p, 1)
+        env = dribble(self.ip, self.server_p - 1, self.monitor_p, self.robot_type, True)
+        model = GAIL.load(args["model_file"], env=env)
+
+        try:
+            self.export_model(args["model_file"], args["model_file"] + ".pkl",
+                              False)  # Export to pkl to create custom behavior
+            self.test_model(model, env, log_path=args["folder_dir"], model_path=args["folder_dir"])
+        except KeyboardInterrupt:
+            print()
+
+        env.close()
+        server.kill()

world/commons/Path_Manager.py

@@ -94,7 +94,9 @@ class Path_Manager():
 
             def get_hard_radius(t):
                 if t.unum in priority_unums:
-                    return 1.0 # extra distance for priority roles
+                    if len(priority_unums) > 2:
+                        return 1.5
+                    return None
                 else:
                     return t.state_ground_area[1]+0.2
 
@@ -113,8 +115,8 @@ class Path_Manager():
                 soft_radius = 0.6
                 hard_radius = lambda o : 0.2
             elif mode == Path_Manager.MODE_DRIBBLE:
-                soft_radius = 2.3
-                hard_radius = lambda o : o.state_ground_area[1]+0.9
+                soft_radius = 1.6
+                hard_radius = lambda o : o.state_ground_area[1]+0.3
             else:
                 soft_radius = 1.0
                 hard_radius = lambda o : o.state_ground_area[1]+0.2