Sovereign 8s.is mirror of standardgalactic/PhyFilter
  • Python 92.1%
  • MATLAB 7.9%
Find a file
Repository files (latest commit first)
Filename Latest commit message Latest commit date
2025-12-22 17:19:18 +08:00
acceleration_case Update all READMEs and add quadruped_case 2025-12-21 19:32:11 +08:00
aerial_manipulation_case Update all READMEs and add quadruped_case 2025-12-21 19:32:11 +08:00
auto_learning Update all READMEs and add quadruped_case 2025-12-21 19:32:11 +08:00
drone_case Update all READMEs and add quadruped_case 2025-12-21 19:32:11 +08:00
quadruped_case Upload quadruped_case module 2025-12-21 19:27:03 +08:00
README.md Update README with quadruped case details 2025-12-22 17:19:18 +08:00

PhyFilter Project Repository

This repository contains source code and experimental data for the PhyFilter (Physics-Informed Filter) research project. The codebase is organized into five main modules, each addressing a specific application or theoretical aspect:

Directory Structure

1. acceleration_case

Acceleration Perception Simulation A case study on acceleration perception using PhyFilter.

  • Stack: MATLAB/Simulink.
  • Features: Neural network training pipeline, simulation models for comparative analysis (vs. Tracking Differentiator, Robust Differentiator), and ablation study scripts.

2. quadruped_case

Quadruped Robot Reinforcement Learning Implements a reinforcement learning framework for quadrupedal locomotion using PhyFilter.

  • Stack: Python, Isaac Gym.
  • Features: Complete training pipeline, simulation environments, and comparative benchmarks against baseline policies.

3. drone_case

Drone Flight Control Provides experimental code for drone state estimation and control.

  • Stack: MATLAB/Simulink.
  • Features: Implements PhyFilter and SEER-I (benchmark) algorithms. Contains the core main.slx model ready for deployment on flight platforms.

4. aerial_manipulation_case

Aerial Manipulation Control Focuses on control strategies for aerial manipulation systems.

  • Stack: MATLAB/Simulink.
  • Features: Includes offline parameter training, a PhyFilter-based dynamic controller, and simulation models (AM_positional_loop.slx, AM_rotational_loop.slx) ready for deployment on flight platforms.

5. auto_learning

Gradient-based Parameter Optimization An automated framework for estimating filter parameters (specifically the pole) of dynamic systems.

  • Stack: MATLAB.
  • Features: Integrates a pre-trained Feedforward Neural Network (FNN) with classical control theory using adjoint-based gradient optimization.

Getting Started

Detailed instructions, dependencies, and file descriptions are provided in the README.md file within each subdirectory.