Downloads · 30 days
0
qads-core
Premchan369/qads-core
qads-core is a reinforcement learning model from Premchan369. Use it for the reinforcement learning task on the model card, and read the license before you ship it in a product. It is set up for pennylane. The card lists the license as mit.
TL;DR: A robot brain that thinks quantum — only when the world gets confusing.
Access
Public
Updated May 10, 2026
Repo size
—
Likes
0
Public
Repo makeup
Click a slice to open those files.
.py111 KB · 88%
At a glance
- Task
- Reinforcement Learning
- Library
- pennylane
- License
- mit
- Access
- Public
- Created
- May 8, 2026
- Updated
- May 10, 2026
- SHA
- 1a23d4d3
Model card
From the Hugging Face model README
⚛️ QADS — Quantum Autonomous Decision System
TL;DR: A robot brain that thinks quantum — only when the world gets confusing.
🤔 What Is This?
Imagine driving through fog at night. Your GPS says "turn left here." But there could be a wall, a person, or nothing — you don't know.
A normal robot planner trusts the GPS and crashes. QADS doesn't. It holds ALL possibilities in its head at once — like a chess master seeing every move ahead — and picks the safest path, not just the shortest.
🧠 How It Works
Sensors see the world
↓
Build a danger map (walls = red, fog = yellow)
↓
Smart Switch: "Is this chaotic?"
↓
YES → Quantum Engine (checks 100 paths at once)
NO → Classical Planner (fastest route)
↓
Robot follows the safest path
📊 The Numbers
Tested hundreds of times from easy rooms to chaotic mazes.
| Difficulty | Classical Wins? | Quantum Wins? | Improvement |
|---|---|---|---|
| 🟢 Easy (clear room) | 85% | 92% | +7% |
| 🟡 Medium (some fog) | 48% | 71% | +23% |
| 🔴 Hard (maze + thick fog) | 18% | 43% | +25% |
In hard conditions, classical crashes 4 of 5 times. Quantum cuts failure in half.
| Metric | Classical | Quantum | Winner |
|---|---|---|---|
| Steps to goal (hard) | 68.5 | 49.3 | 28% faster |
| Collisions (hard) | 4.2 | 2.7 | 36% fewer crashes |
| Decision speed | 50ms | 450ms | Classical faster but both real-time |
🔑 Smart Switch
Quantum only fires when needed:
- Easy world → 12% quantum usage
- Hard world → 89% quantum usage
- Average → 46% usage
You pay for quantum only when it actually helps.
🎮 Live Demo
Three tabs: Generate Environment, Run Mission, Run Benchmark.
📦 What's Inside
| Module | What It Does |
|---|---|
| Quantum Core | Think about many paths at once |
| Hybrid Planner | Smart switch: classical or quantum |
| World Graph | Danger map with risk scores |
| A* Planner | Classic shortest-path |
| RL Agent | Learns from mistakes |
| K2 Think | Explains why it chose a path |
🌍 Real-World Use
| Problem | QADS Fix |
|---|---|
| 🚁 Drone in storm | Avoids wind, not just shortest |
| 🏭 100 warehouse robots | Schedules without jams |
| 🚗 Self-driving in fog | Plans for "maybe a pedestrian" |
| 🛰️ Mars rover | Pre-computes backup routes |
🏁 One Sentence
QADS is a robot brain that thinks quantum when the world gets confusing — not all the time — cutting crashes in half when normal planners give up.
Repo: Premchan369/qads-core
Demo: QADS Live
License: MIT
- Task
- Reinforcement Learning
- Library
- pennylane
- License
- mit
- Languages
- en
- Created
- May 8, 2026
- Updated
- May 10, 2026