Merge 9496d0f into fc115c5

CHANGELOG.md

@@ -6,8 +6,13 @@ All notable changes to this project will be documented in this file.
 
 ### Added
 
+- Add ``upkie_locomotion.envs.register`` function
 - PPO balancer: setting for total number of training timesteps
 
+### Changed
+
+- UpkieWheelsEnv: remove dependency on gin
+
 ## [0.2.0] - 2023/03/03
 
 ### Added

agents/ppo_balancer/test.py

@@ -11,10 +11,10 @@
 import gin
 import mpacklog
 from loop_rate_limiters import AsyncRateLimiter
+from settings import Settings
 from stable_baselines3 import PPO
-from upkie_locomotion.envs import UpkieWheelsEnv
 
-from settings import Settings
+from upkie_locomotion.envs import UpkieWheelsEnv
 
 keep_going = True
 
@@ -60,7 +60,6 @@ def load_policy(agent_dir: str, policy_name: str):
 
 if __name__ == "__main__":
     agent_dir = os.path.abspath(os.path.dirname(__file__))
-    gin.parse_config_file(UpkieWheelsEnv.gin_config())
     gin.parse_config_file(f"{agent_dir}/settings.gin")
 
     parser = argparse.ArgumentParser(description=__doc__)

agents/ppo_balancer/train.py

@@ -187,7 +187,6 @@ def get_bullet_argv(agent_name: str, show: bool) -> List[str]:
     args = parser.parse_args()
 
     agent_dir = os.path.dirname(__file__)
-    gin.parse_config_file(UpkieWheelsEnv.gin_config())
     gin.parse_config_file(f"{agent_dir}/settings.gin")
 
     agent_name = generate_agent_name()

envs/BUILD

@@ -15,7 +15,6 @@ py_library(
     ],
     data = [
         "spine.yaml",
-        "upkie_wheels_env.gin",
     ],
     deps = [
         "@upkie_locomotion//observers/base_pitch",

envs/__init__.py

@@ -15,8 +15,20 @@
 # See the License for the specific language governing permissions and
 # limitations under the License.
 
+import gym
+
 from .upkie_wheels_env import UpkieWheelsEnv
 
+
+def register():
+    gym.envs.register(
+        id=f"UpkieWheelsEnv-v{UpkieWheelsEnv.version}",
+        entry_point="upkie_locomotion.envs:UpkieWheelsEnv",
+        max_episode_steps=1_000_000_000,
+    )
+
+
 __all__ = [
     "UpkieWheelsEnv",
+    "register",
 ]

envs/tests/upkie_wheels_env_test.py

@@ -19,7 +19,6 @@
 
 import unittest
 
-import gin
 import numpy as np
 import posix_ipc
 
@@ -61,7 +60,6 @@ def get_observation(self) -> dict:
 
 class TestUpkieWheelsEnv(unittest.TestCase):
     def setUp(self):
-        gin.parse_config_file(UpkieWheelsEnv.gin_config())
         shm_name = "/vroum"
         shared_memory = posix_ipc.SharedMemory(
             shm_name, posix_ipc.O_RDWR | posix_ipc.O_CREAT, size=42

envs/upkie_wheels_env.py

@@ -20,7 +20,6 @@
 from os import path
 from typing import Dict, Optional, Tuple, Union
 
-import gin
 import gym
 import numpy as np
 import yaml
@@ -36,7 +35,6 @@
 MAX_IMU_ANGULAR_VELOCITY: float = 1000.0  # rad/s
 
 
-@gin.configurable
 class UpkieWheelsEnv(gym.Env):
 
     """!
@@ -109,22 +107,13 @@ class UpkieWheelsEnv(gym.Env):
         for joint in ("hip", "knee")
     ]
 
-    def id(self) -> str:
-        """!
-        Name and version of this environment for registration.
-
-        Returns:
-            Name and version of the environment.
-        """
-        return f"UpkieWheelsEnv-v{self.version}"
-
     def __init__(
         self,
-        config: Optional[dict],
-        fall_pitch: float,
-        max_ground_velocity: float,
-        shm_name: str,
-        wheel_radius: float,
+        config: Optional[dict] = None,
+        fall_pitch: float = 1.0,
+        max_ground_velocity: float = 1.0,
+        shm_name: str = "/vulp",
+        wheel_radius: float = 0.06,
     ):
         """!
         Initialize environment.
@@ -210,15 +199,6 @@ def _observe(self) -> dict:
         self.last_observation = observation_dict
         return observation_dict
 
-    @staticmethod
-    def gin_config():
-        """!
-        Path to the Gin configuration for this environment.
-        """
-        dirname = path.dirname(__file__)
-        basename = path.basename(__file__).replace(".py", ".gin")
-        return f"{dirname}/{basename}"
-
     def vectorize_observation(self, observation_dict: dict) -> np.ndarray:
         """!
         Extract observation vector from a full observation dictionary.

envs/upkie_wheels_reward.py

@@ -5,11 +5,9 @@
 
 from typing import Tuple
 
-import gin
 import numpy as np
 
 
-@gin.configurable
 class UpkieWheelsReward:
 
     """!
@@ -42,11 +40,11 @@ def get_range() -> Tuple[float, float]:
 
     def __init__(
         self,
-        lookahead_duration: float,
-        max_pitch: float,
-        max_position: float,
-        pitch_weight: float,
-        position_weight: float,
+        lookahead_duration: float = 0.1,
+        max_pitch: float = 1.5707963267948966,
+        max_position: float = 0.5,
+        pitch_weight: float = 1.0,
+        position_weight: float = 1.0,
     ):
         """!
         Initialize reward.

examples/upkie_wheels_env_simple.py

@@ -0,0 +1,24 @@
+#!/usr/bin/env python3
+# -*- coding: utf-8 -*-
+#
+# Copyright 2023 Inria
+
+import gym
+import numpy as np
+
+import upkie_locomotion.envs
+
+if __name__ == "__main__":
+    upkie_locomotion.envs.register()
+    env = gym.make("UpkieWheelsEnv-v1")
+    observation = env.reset(seed=42)
+
+    action = np.zeros(env.action_space.shape)
+    for step in range(1_000_000):
+        observation, reward, done, _ = env.step(action)
+        if done:
+            observation = env.reset()
+        pitch = observation[0]
+        action[0] = 10.0 * pitch
+
+    env.close()

pypi/tmpflit.sh → tools/pypi/tmpflit.sh

@@ -27,7 +27,7 @@ fi
 
 BASEDIR=$(dirname 0)
 COMMAND=$@
-SRCDIR=${BASEDIR}/..
+SRCDIR=${BASEDIR}/../..
 TMPDIR=$(mktemp -d)
 
 echo "[debug] COMMAND=${COMMAND}"

pypi/upkie_locomotion/__init__.py → tools/pypi/upkie_locomotion/__init__.py

@@ -17,4 +17,4 @@
 
 """Real-time motion control for Python."""
 
-__version__ = "0.3.0rc0"
+__version__ = "0.3.0rc2"