Clean up android_env configuration, packaging, and module structure:

README.md

@@ -158,7 +158,8 @@ $ bazel run //android_env/playground:playground -- \
 
 ### 4. Reinforcement Learning (`dm_env`)
 
-To define MDP tasks with rewards and episode resets, see the
+To define MDP tasks with rewards and episode resets (after
+`pip install "android-env[rl]"`), see the
 [Tasks guide](https://github.com/google-deepmind/android_env/blob/gh-pages/docs/tasks_guide.md),
 [Example tasks](https://github.com/google-deepmind/android_env/blob/gh-pages/docs/example_tasks.md),
 and

README.zh.md

@@ -120,7 +120,8 @@ $ bazel run //android_env/playground:playground -- \
 
 ### 4. 强化学习接入 (`dm_env`)
 
-如需定义包含奖励机制与回合重置的 MDP 任务，请参阅[任务指南](https://github.com/google-deepmind/android_env/blob/gh-pages/docs/tasks_guide.md)、[示例任务库](https://github.com/google-deepmind/android_env/blob/gh-pages/docs/example_tasks.md)以及[详细操作说明](https://github.com/google-deepmind/android_env/blob/gh-pages/docs/instructions.md)。
+如需定义包含奖励机制与回合重置的 MDP 任务（通过 `pip install "android-env[rl]"`
+安装后），请参阅[任务指南](https://github.com/google-deepmind/android_env/blob/gh-pages/docs/tasks_guide.md)、[示例任务库](https://github.com/google-deepmind/android_env/blob/gh-pages/docs/example_tasks.md)以及[详细操作说明](https://github.com/google-deepmind/android_env/blob/gh-pages/docs/instructions.md)。
 
 ## 关于 AndroidEnv (About)
 

android_env/BUILD.bazel

@@ -3,35 +3,6 @@ load("@rules_python//python:packaging.bzl", "py_package", "py_wheel")
 
 package(default_visibility = ["//visibility:public"])
 
-py_library(
-    name = "android_env",
-    srcs = ["__init__.py"],
-    deps = [
-        ":env_interface",
-        ":loader",
-        "//android_env/components:android_device",
-        "//android_env/components:android_device_loader",
-        "//android_env/components:config_classes",
-        "//android_env/rl:environment",
-    ],
-)
-
-py_test(
-    name = "init_test",
-    srcs = ["init_test.py"],
-    legacy_create_init = False,
-    deps = [
-        ":android_env",
-        ":env_interface",
-        ":loader",
-        "//android_env/components:android_device",
-        "//android_env/components:android_device_loader",
-        "//android_env/components:config_classes",
-        "//android_env/rl:environment",
-        "@pip//absl_py",
-    ],
-)
-
 py_library(
     name = "env_interface",
     srcs = ["env_interface.py"],
@@ -78,7 +49,6 @@ py_package(
     name = "android_env_pkg",
     packages = ["android_env"],
     deps = [
-        ":android_env",
         ":loader",
         "//android_env/components:android_device_loader",
         "//android_env/mcp:android_mcp_server_lib",
@@ -112,16 +82,19 @@ py_wheel(
         ],
     },
     extra_requires = {
-        "gym": ["gymnasium>=0.29.1"],
+        "gym": [
+            "dm-env>=1.6",
+            "gymnasium>=0.29.1",
+        ],
         "mcp": ["mcp>=1.26,<2"],
         "playground": ["ml-collections"],
+        "rl": ["dm-env>=1.6"],
     },
     homepage = "https://github.com/google-deepmind/android_env",
     license = "Apache License, Version 2.0",
     python_requires = ">=3.10",
     requires = [
         "absl-py>=2.1.0",
-        "dm-env>=1.6",
         "numpy>=2.1.0",
         "portpicker>=1.6.0",
         "pillow>=10.4.0",

android_env/components/BUILD.bazel

@@ -2,11 +2,6 @@ load("@rules_python//python:defs.bzl", "py_library", "py_test")
 
 package(default_visibility = ["//visibility:public"])
 
-py_library(
-    name = "action_type",
-    srcs = ["action_type.py"],
-)
-
 py_library(
     name = "adb_call_parser",
     srcs = ["adb_call_parser.py"],

android_env/components/adb_controller.py

@@ -72,7 +72,7 @@ def init_server(self, timeout: float | None = None):
     """Initialize the ADB server deamon on the given port.
 
     This function should be called immediately after initializing the first
-    adb_controller, and before launching the simulator.
+    adb_controller, and before launching the emulator or connecting to a device.
 
     Args:
       timeout: A timeout to use for this operation. If not set the default

android_env/components/config_classes.py

@@ -176,6 +176,6 @@ class AndroidEnvConfig:
   task_manager: TaskManagerConfig = dataclasses.field(
       default_factory=TaskManagerConfig
   )
-  simulator: AndroidDeviceConfig = dataclasses.field(
+  device: AndroidDeviceConfig = dataclasses.field(
       default_factory=AndroidDeviceConfig
   )

android_env/components/errors.py

@@ -25,35 +25,17 @@ class AndroidEnvError(Exception):
 
 
 class ReadObservationError(AndroidEnvError):
-  """When the environment is unable to obtain an observation from a simulator."""
+  """When the environment is unable to obtain an observation from the device."""
 
   ERROR_CODE = 1
 
 
-class CoordinatorError(AndroidEnvError):
-  """Error raised by the Coordinator."""
-
-  ERROR_CODE = 2
-
-
-class TooManyRestartsError(CoordinatorError):
-  """The number of restarts has exceeded _MAX_RESTART_TRIES."""
-
-  ERROR_CODE = 3
-
-
 class AdbControllerError(AndroidEnvError):
   """Errors that can be raised by ADBController."""
 
   ERROR_CODE = 4
 
 
-class SimulatorError(AndroidEnvError):
-  """Errors that can be raised by a simulator."""
-
-  ERROR_CODE = 5
-
-
 class SendActionError(AndroidEnvError):
   """Raised when action couldn't be sent successfully."""
 
@@ -90,10 +72,7 @@ def from_code(code: int, msg: str = '') -> AndroidEnvError | None:
   code_to_error = {
       0: AndroidEnvError,
       1: ReadObservationError,
-      2: CoordinatorError,
-      3: TooManyRestartsError,
       4: AdbControllerError,
-      5: SimulatorError,
       6: SendActionError,
       7: StepCommandError,
       8: WaitForAppScreenError,

android_env/components/errors_test.py

@@ -24,10 +24,7 @@ class ErrorsTest(parameterized.TestCase):
 
   @parameterized.parameters(
       (errors.ReadObservationError, 1),
-      (errors.CoordinatorError, 2),
-      (errors.TooManyRestartsError, 3),
       (errors.AdbControllerError, 4),
-      (errors.SimulatorError, 5),
       (errors.SendActionError, 6),
       (errors.StepCommandError, 7),
       (errors.WaitForAppScreenError, 8),
@@ -42,10 +39,7 @@ def test_error_codes_unique(self):
     error_codes = set()
     errors_list = (
         errors.ReadObservationError,
-        errors.CoordinatorError,
-        errors.TooManyRestartsError,
         errors.AdbControllerError,
-        errors.SimulatorError,
         errors.SendActionError,
         errors.StepCommandError,
         errors.WaitForAppScreenError,
@@ -57,10 +51,7 @@ def test_error_codes_unique(self):
 
   @parameterized.parameters([
       errors.ReadObservationError(),
-      errors.CoordinatorError(),
-      errors.TooManyRestartsError(),
       errors.AdbControllerError(),
-      errors.SimulatorError(),
       errors.SendActionError(),
       errors.StepCommandError(),
       errors.WaitForAppScreenError(),
@@ -86,10 +77,7 @@ def test_from_code_unsupported_code(self, code: int):
       (0, errors.AndroidEnvError, 'hello'),
       (0, errors.AndroidEnvError, ''),
       (1, errors.ReadObservationError, 'Could not read obs.'),
-      (2, errors.CoordinatorError, 'Some error'),
-      (3, errors.TooManyRestartsError, 'Too many already...'),
       (4, errors.AdbControllerError, 'Some adb error...'),
-      (5, errors.SimulatorError, 'Simulator is not coping.'),
       (6, errors.SendActionError, 'Could not send action.'),
       (7, errors.StepCommandError, 'Some issue setting up the task.'),
       (8, errors.WaitForAppScreenError, 'Waited for too long!'),

android_env/loader.py

@@ -46,12 +46,12 @@ def load(
 ) -> env_interface.AndroidEnvInterface:
   """Loads an AndroidEnv instance."""
 
-  if isinstance(config.simulator, config_classes.AndroidDeviceConfig):
-    device = android_device_loader.load_android_device(config.simulator)
+  if isinstance(config.device, config_classes.AndroidDeviceConfig):
+    device = android_device_loader.load_android_device(config.device)
     env = rl_environment.AndroidEnv(device=device)
     if config.task != config_classes.TaskConfig():
       task = _load_task(config.task)
       env = task_wrapper.TaskWrapper(env, task=task, config=config.task_manager)
     return env
 
-  raise ValueError(f'Unsupported simulator config: {config.simulator}')
+  raise ValueError(f'Unsupported device config: {config.device}')