Information-Entropic Navigation (IEN): Reduced-Order Simulation

🚀 Overview

This repository contains the reference implementation and validation experiments for the Information-Entropic Navigation (IEN) protocol.

We present a reduced-order dynamical simulation demonstrating the emergent control behavior predicted by Information-Entropic Navigation. While the full Active Inference loop is abstracted, the resulting continuous low-thrust control law shows that early correction of state divergence using environmental forces eliminates the need for impulsive Δv in station-keeping scenarios.

The core hypothesis is that by minimizing the Kullback-Leibler divergence ($D_{KL}$) between the internal model and the sensor state, a spacecraft can maintain phase coherence using only Solar Radiation Pressure (SRP), effectively reducing station-keeping propellant costs to near zero.

---

📊 The "Kill Shot": Efficiency Comparison

The following simulation compares a classical Deadband Control (Chemical) agent against the IEN Agent (SRP) under identical environmental disturbances (solar wind stochastic drift).

(Figure 1: Top - Position holding capability. Bottom - Cumulative Delta-V expenditure. Note the IEN agent maintains position with zero propellant usage.)

---

⚙️ Simulation Scope & Limitations

This is a Proof-of-Concept (PoC) implementation designed to validate the control law dynamics. Reviewers should note the following design choices:

1. 1D Longitudinal Dynamics: The simulation models a 1-degree-of-freedom "Station Keeping" scenario (e.g., maintaining distance along the sun-line in a Halo Orbit). It is a "Toy Model" for behavioral analysis.
2. Abstracted Inference: The IENAgent assumes an idealized observer where the belief update is instantaneous. The explicit Bayesian Filter (UKF/Particle Filter) is implicit in the control gain derivation for this version.
3. SRP Availability: We assume a constant solar aspect angle capability. In a full 6-DOF implementation, attitude maneuvers would be coupled with power generation constraints.
4. Gain Tuning: Control gains ($K_p, K_v$) are tuned for stability in this reduced order model, rather than dynamically derived from the $\nabla D_{KL}$ manifold in real-time.

---

🛠️ Usage

Prerequisites

• Python 3.8+
• NumPy
• Matplotlib

Running the Comparison

python simulation_compare.py

🔮 Future Work (Phase 2)

Extension to CR3BP (Circular Restricted Three-Body Problem) dynamics for L1/L2 Halo Orbits (2D/3D).

Implementation of an explicit Unscented Kalman Filter (UKF) to drive the Active Inference loop.

Coupling with ADCS (Attitude Determination and Control Systems) constraints.

📜 Citation If you use this logic in your research, please cite the foundational paper:

Brito, L. (2025). Information-Entropic Navigation (IEN): Active Inference for Low-Thrust Trajectory Optimization. Zenodo. https://doi.org/10.5281/zenodo.17930558

Maintained by Leonel Brito | Researcher in Evolutionary Systems Dynamics & Information Physics.