File size: 3,526 Bytes
9cb40fe | 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 | # Copyright (c) Space Robotics Lab, SnT, University of Luxembourg, SpaceR
# RANS: Reinforcement Learning based Autonomous Navigation for Spacecrafts
# arXiv:2310.07393 — El-Hariry, Richard, Olivares-Mendez
#
# OpenEnv-compatible implementation
"""Base class for RANS spacecraft navigation tasks."""
from __future__ import annotations
import math
from abc import ABC, abstractmethod
from typing import Any, Dict, Tuple
import numpy as np
class BaseTask(ABC):
"""
Abstract base class for all RANS spacecraft navigation tasks.
Subclasses define:
- The task-specific observation vector
- The reward function (matching the RANS paper's formulations)
- Target generation and episode reset logic
"""
def __init__(self, config: Dict[str, Any] | None = None) -> None:
self.config: Dict[str, Any] = config or {}
self._target: Any = None
# ------------------------------------------------------------------
# Abstract interface
# ------------------------------------------------------------------
@abstractmethod
def reset(self, spacecraft_state: np.ndarray) -> Dict[str, Any]:
"""
Sample a new target and reset internal episode state.
Args:
spacecraft_state: Current state vector [x, y, θ, vx, vy, ω].
Returns:
Dictionary with task metadata (target values, etc.).
"""
@abstractmethod
def get_observation(self, spacecraft_state: np.ndarray) -> np.ndarray:
"""
Compute the task-specific observation vector from the spacecraft state.
Args:
spacecraft_state: Current state [x, y, θ, vx, vy, ω].
Returns:
1-D float32 array of length ``num_observations``.
"""
@abstractmethod
def compute_reward(
self, spacecraft_state: np.ndarray
) -> Tuple[float, bool, Dict[str, Any]]:
"""
Compute reward, done flag, and diagnostic info.
Args:
spacecraft_state: Current state [x, y, θ, vx, vy, ω].
Returns:
(reward, done, info) tuple.
"""
# ------------------------------------------------------------------
# Common helpers
# ------------------------------------------------------------------
@property
def num_observations(self) -> int:
"""Size of the task-specific state observation vector."""
return 0
@property
def name(self) -> str:
return self.__class__.__name__
# ------------------------------------------------------------------
# Shared reward primitives (from RANS paper Sec. IV-B)
# ------------------------------------------------------------------
@staticmethod
def _reward_exponential(error: float, sigma: float) -> float:
"""exp(-error² / (2·σ²)) — tight peak near zero."""
return math.exp(-(error**2) / max(2.0 * sigma**2, 1e-9))
@staticmethod
def _reward_inverse(error: float) -> float:
"""1 / (1 + error) — smooth monotone decay."""
return 1.0 / (1.0 + error)
@staticmethod
def _wrap_angle(angle: float) -> float:
"""Wrap angle to (−π, π]."""
return (angle + math.pi) % (2.0 * math.pi) - math.pi
@staticmethod
def _world_to_body(dx: float, dy: float, theta: float) -> Tuple[float, float]:
"""Rotate world-frame displacement into body frame."""
c, s = math.cos(theta), math.sin(theta)
return c * dx + s * dy, -s * dx + c * dy
|