File size: 10,661 Bytes
7078c69 5e3ab5d 7078c69 5e3ab5d 7078c69 5e3ab5d 7078c69 5e3ab5d 7078c69 5e3ab5d 7078c69 | 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60 61 62 63 64 65 66 67 68 69 70 71 72 73 74 75 76 77 78 79 80 81 82 83 84 85 86 87 88 89 90 91 92 93 94 95 96 97 98 99 100 101 102 103 104 105 106 107 108 109 110 111 112 113 114 115 116 117 118 119 120 121 122 123 124 125 126 127 128 129 130 131 132 133 134 135 136 137 138 139 140 141 142 143 144 145 146 147 148 149 150 151 152 153 154 155 156 157 158 159 160 161 162 163 164 165 166 167 168 169 170 171 172 173 174 175 176 177 178 179 180 181 182 183 184 185 186 187 188 189 190 191 192 193 194 195 196 197 198 199 200 201 202 203 204 205 206 207 208 209 210 211 212 213 214 215 216 217 218 219 220 221 222 223 224 225 226 227 228 229 230 231 232 233 234 235 236 237 238 239 240 241 242 243 244 245 246 247 248 249 250 251 252 253 254 255 256 257 258 259 260 261 262 263 264 265 266 267 268 | # Copyright (c) Space Robotics Lab, SnT, University of Luxembourg, SpaceR
# RANS: arXiv:2310.07393 β OpenEnv-compatible implementation
"""
RANSEnvironment
===============
OpenEnv ``Environment`` subclass that wraps the 2-D spacecraft simulator and
the RANS task suite.
Supported tasks (set via RANS_TASK env-var or constructor argument):
β’ GoToPosition β reach a target (x, y)
β’ GoToPose β reach a target (x, y, ΞΈ)
β’ TrackLinearVelocity β maintain (vx_t, vy_t)
β’ TrackLinearAngularVelocity β maintain (vx_t, vy_t, Ο_t)
The environment follows the RANS paper (arXiv:2310.07393) physics and reward
formulations, adapted to run in CPU-only Docker containers without Isaac Gym.
"""
from __future__ import annotations
import math
import os
import uuid
from typing import Any, Dict, Optional
import numpy as np
try:
from openenv.core.env_server.interfaces import Action, Environment, Observation
except ImportError:
from pydantic import BaseModel as Action # type: ignore[assignment]
from pydantic import BaseModel as Environment # type: ignore[assignment]
from pydantic import BaseModel as Observation # type: ignore[assignment]
try:
# Installed package import
from rans_env.models import SpacecraftAction, SpacecraftObservation, SpacecraftState
from rans_env.server.spacecraft_physics import Spacecraft2D, SpacecraftConfig
from rans_env.server.tasks import TASK_REGISTRY
except ImportError:
# Development / test import (package not yet installed, RANS dir on sys.path)
import sys, os as _os
sys.path.insert(0, _os.path.dirname(_os.path.dirname(__file__)))
from models import SpacecraftAction, SpacecraftObservation, SpacecraftState # type: ignore[no-redef]
from server.spacecraft_physics import Spacecraft2D, SpacecraftConfig # type: ignore[no-redef]
from server.tasks import TASK_REGISTRY # type: ignore[no-redef]
class RANSEnvironment(Environment):
"""
RANS spacecraft navigation environment for OpenEnv.
References
----------
El-Hariry, Richard, Olivares-Mendez (2023).
"RANS: Highly-Parallelised Simulator for Reinforcement Learning based
Autonomous Navigating Spacecrafts." arXiv:2310.07393.
"""
def __init__(
self,
task: str = "GoToPosition",
spacecraft_config: Optional[SpacecraftConfig] = None,
task_config: Optional[Dict[str, Any]] = None,
max_episode_steps: int = 500,
initial_pos_range: float = 2.0,
initial_vel_range: float = 0.1,
) -> None:
"""
Parameters
----------
task:
One of TASK_REGISTRY keys. Overridden by RANS_TASK env-var.
spacecraft_config:
Physical platform configuration. Uses 8-thruster MFP2D default.
task_config:
Dict of task hyper-parameters forwarded to the task constructor.
max_episode_steps:
Hard step limit per episode (overrides RANS_MAX_STEPS env-var).
initial_pos_range:
Half-width of the uniform distribution for random initial position.
initial_vel_range:
Half-width for random initial velocities.
"""
# Allow env-var overrides (useful for Docker deployments)
task = os.environ.get("RANS_TASK", task)
max_episode_steps = int(
os.environ.get("RANS_MAX_STEPS", str(max_episode_steps))
)
if task not in TASK_REGISTRY:
raise ValueError(
f"Unknown task '{task}'. "
f"Available: {sorted(TASK_REGISTRY.keys())}"
)
self._task_name = task
self._max_episode_steps = max_episode_steps
self._initial_pos_range = initial_pos_range
self._initial_vel_range = initial_vel_range
# Physics simulator β RANS_NUM_THRUSTERS overrides spacecraft_config
if spacecraft_config is None:
n_env = os.environ.get("RANS_NUM_THRUSTERS")
if n_env is not None:
n = int(n_env)
spacecraft_config = SpacecraftConfig.from_num_thrusters(n)
else:
spacecraft_config = SpacecraftConfig.default_8_thruster()
self._spacecraft = Spacecraft2D(spacecraft_config)
# Task
self._task = TASK_REGISTRY[task](task_config or {})
# Episode bookkeeping
self._step_count: int = 0
self._total_reward: float = 0.0
self._ep_state = SpacecraftState(task=self._task_name)
# ------------------------------------------------------------------
# OpenEnv Environment interface
# ------------------------------------------------------------------
def reset(self) -> Observation:
"""Start a new episode with a randomised initial spacecraft state."""
init_state = self._sample_initial_state()
self._spacecraft.reset(init_state)
task_info = self._task.reset(self._spacecraft.state)
self._step_count = 0
self._total_reward = 0.0
self._ep_state = SpacecraftState(
episode_id=str(uuid.uuid4()),
step_count=0,
task=self._task_name,
**self._physical_state_dict(),
)
return self._make_observation(reward=0.0, done=False, info=task_info)
def step(self, action: Action) -> Observation:
"""
Advance the simulation by one step.
Control mode is selected automatically based on which fields are set:
1. **Thrusters** (``action.thrusters`` is not None):
List of per-thruster activations β [0, 1].
2. **Force/torque** (``action.fx``, ``action.fy``, or ``action.torque``
is not None):
Direct world-frame force/torque β bypasses thruster geometry.
3. **Velocity target** (``action.vx_target``, ``action.vy_target``, or
``action.omega_target`` is not None):
PD controller drives the spacecraft toward the requested velocities.
"""
if not hasattr(action, "thrusters"):
raise ValueError(
f"Expected SpacecraftAction, received {type(action).__name__}."
)
# ββ Mode 1: thruster activations βββββββββββββββββββββββββββββββββ
if action.thrusters is not None:
activations = np.array(action.thrusters, dtype=np.float64)
n = self._spacecraft.n_thrusters
if len(activations) != n:
padded = np.zeros(n, dtype=np.float64)
padded[: min(len(activations), n)] = activations[:n]
activations = padded
self._spacecraft.step(activations)
# ββ Mode 2: direct force / torque ββββββββββββββββββββββββββββββββ
elif any(v is not None for v in [action.fx, action.fy, action.torque]):
self._spacecraft.step_force_torque(
fx_world=float(action.fx or 0.0),
fy_world=float(action.fy or 0.0),
torque=float(action.torque or 0.0),
)
# ββ Mode 3: velocity target βββββββββββββββββββββββββββββββββββββββ
elif any(
v is not None
for v in [action.vx_target, action.vy_target, action.omega_target]
):
self._spacecraft.step_velocity_target(
vx_target=float(action.vx_target or 0.0),
vy_target=float(action.vy_target or 0.0),
omega_target=float(action.omega_target or 0.0),
)
# ββ No action β advance with zero forces ββββββββββββββββββββββββββ
else:
self._spacecraft.step(np.zeros(self._spacecraft.n_thrusters))
self._step_count += 1
# Compute task reward
reward, goal_reached, info = self._task.compute_reward(
self._spacecraft.state
)
self._total_reward += reward
# Determine episode termination
done = goal_reached or (self._step_count >= self._max_episode_steps)
# Rebuild persistent state (Pydantic models are immutable by default)
self._ep_state = SpacecraftState(
episode_id=self._ep_state.episode_id,
step_count=self._step_count,
task=self._task_name,
total_reward=self._total_reward,
goal_reached=goal_reached,
**self._physical_state_dict(),
)
return self._make_observation(reward=reward, done=done, info=info)
@property
def state(self) -> SpacecraftState:
return self._ep_state
# ------------------------------------------------------------------
# Helpers
# ------------------------------------------------------------------
def _sample_initial_state(self) -> np.ndarray:
"""Uniform random initial state (small velocities, random pose)."""
r = self._initial_pos_range
v = self._initial_vel_range
return np.array(
[
np.random.uniform(-r, r), # x
np.random.uniform(-r, r), # y
np.random.uniform(-math.pi, math.pi), # ΞΈ
np.random.uniform(-v, v), # vx
np.random.uniform(-v, v), # vy
np.random.uniform(-v, v), # Ο
],
dtype=np.float64,
)
def _physical_state_dict(self) -> Dict[str, float]:
s = self._spacecraft.state
return {
"x": float(s[0]),
"y": float(s[1]),
"heading_rad": float(s[2]),
"vx": float(s[3]),
"vy": float(s[4]),
"angular_velocity_rads": float(s[5]),
}
def _make_observation(
self, reward: float, done: bool, info: Dict[str, Any]
) -> SpacecraftObservation:
task_obs = self._task.get_observation(self._spacecraft.state)
return SpacecraftObservation(
state_obs=task_obs.tolist(),
thruster_transforms=self._spacecraft.get_thruster_transforms().tolist(),
thruster_masks=self._spacecraft.get_thruster_masks().tolist(),
mass=self._spacecraft.config.mass,
inertia=self._spacecraft.config.inertia,
task=self._task_name,
reward=float(reward),
done=bool(done),
info={**info, "step": self._step_count},
)
|