User repository on GitHub
Camera-observed UAV pursuit in MuJoCo with Jev decisions, RGB/depth HUD, replay and episode collection.
- Forks
- 0
- Last pushed
- Updated Oct 2026
Repo stats from the GitHub API, cached Oct 2026.
Jev directory
CommunityJev Demos & Experiments
Multi-UAV JEV
A MuJoCo city simulation where one, two or four camera-observed drones pursue a target: each role asks Jev role-specific typed questions while deterministic mission navigation, safety guards and motor control run independently at 50 Hz.
- Stars
- 54
- Language
- Python
- License
- APACHE-2.0
- Reading time
- 3 min
- Published by
- Community
- Author
- YongLD
- Last verified
- Oct 2026
Highlights
- Jev receives structured text derived from depth, ideal segmentation and odometry; raw RGB/depth tensors are displayed, not submitted to the decision model.
- Role rules live in sim/role.md and reload while running; invalid edits retain the previous configuration.
- P-01 tracks, P-02 blockades ahead, P-03 predicts an interception point, P-04 supports and rescues; 1, 2 or 4 aircraft fly.
- Multi-agent choices older than 0.8 simulated seconds are logged and discarded, and a failed endpoint never becomes a fabricated decision.
- Apache-2.0 for new code with inherited jev-drone code under MIT and CC0 city meshes; 46 stars, created 2026-10-09, pushed 2026-10-13.
Quickstart
bash
git clone https://github.com/YongLD/Multi-UAV-JEV.git
cd Multi-UAV-JEV
bash setup.sh
export JEV_ENDPOINT=http://127.0.0.1:8000/v1/systemone
bash run.sh --drones 4Watch out
README reviewed 2026-10-17; tracking and obstacle avoidance are still experimental, communication is an ideal zero-delay broadcast, the evader is kinematic, and the project warns that favorable GIFs or short checks do not establish improved multi-agent success rates. Model weights and GPU inference are separate from the simulation environment.
Questions about this entry
Short answers drawn from the linked project or page.
A MuJoCo city where one, two or four camera-carrying drones cooperate to pursue a target: P-01 keeps contact, P-02 blockades ahead, P-03 predicts a meeting point, and P-04 supports and rescues. Each aircraft has its own free joint, four motors, camera perception, a controller and a Jev decision worker.
Jev answers role-specific tactical questions over structured text built from depth, ideal MuJoCo segmentation and own odometry; it does not receive pixels. Deterministic navigation, depth braking, peer separation and recovery run independently at 50 Hz, and a busy or failed endpoint never becomes a fabricated decision.
Validated on Linux with Python 3.12, MuJoCo 3.14.0 and EGL: clone the repo, run bash setup.sh, start a compatible NeoHorse endpoint separately, then export JEV_ENDPOINT and run bash run.sh --drones 4. The live cockpit opens at http://127.0.0.1:8080.
Structured text derived from depth, ideal MuJoCo target segmentation and own odometry. Raw RGB and depth tensors are displayed but not submitted, and enemy truth coordinates, building maps and the scripted future route are excluded from navigation and model input, though referee and spectator views may use truth.
The README lists tracking and obstacle avoidance as experimental, communication as an ideal zero-delay broadcast without radio range or packet loss, idealized segmentation, and a kinematic evader. Finite camera coverage and conservative collision hulls can cause contacts, and model or encoder load can slow wall-clock execution.
Recordings keep 1280x720 per aircraft at 20 FPS, and each case directory stores decisions.jsonl, events.jsonl, referee.jsonl, policy.json and summary.json. The project labels these raw training candidates, not expert labels, so model errors, stale responses and controller actions should stay distinguishable.
New contributions are Apache-2.0, inherited jev-drone code keeps MIT terms, and bundled city meshes are CC0. Downloaded robot assets, the NeoHorse-Jev-4B model weights and inference runtimes retain their own licenses, so check those terms before redistributing a full stack.
Reactions & coverage
Posts, threads, and videos about this entry from around the web.
Hacker News: Pick Verify: a single decision at the warehouse pick station catches confusable picks.
Show HN: Verifying robot picks with a decision modelRead the thread on Hacker NewsX: A dual-arm robot with Jev in the middle layer
YouTube: Dan Makes Things drives an ESP32 robot with Jev decisions.
More Jev resources like this
A MuJoCo Franka arm with an Allegro hand stacking a blue cube on a red one, driven two ways: Claude Opus 5 reasoning through low-level tools, and Jev picking each simulated move while code owns the physics.
Jev Demos & Experiments#community#python#robotics
Community
FazalAAli/jev-robotics-demoPython
FazalAAli/jev-robotics-demoPythonAn open-source browser 3D character split between two models: GPT-Live-1 talks with you over WebRTC while Jev reads the live transcript and picks Annie's face, gestures and body language in about 120 ms, with code gates on every answer.
Jev Demos & Experiments#community#github#python
Community
rehan-remade/open-anniePython
rehan-remade/open-anniePythonMuJoCo simulation where a local decider-2b model picks the next skill for an AgileX PiPER arm through a Jev-style typed-choice interface while the red cube target is pushed to new positions mid-run. Deterministic checks may reject a chosen skill but never substitute another.
Jev Demos & Experiments#community#robotics#simulation
Community
Hu-xiao-max/jev_robotPython
Hu-xiao-max/jev_robotPythonAn SO-101 robot pet powered by Jev: send it messages through Telegram or the local interface and it answers with recorded movement and electronic calls, with a choreography mode for arranging motions and sounds.
Jev Tools & CLIs#community#github#robotics
Community
Muurrphy/bipuPython
Muurrphy/bipuPythonA browser game for exploring fast, structured decisions with Laya: play the bird yourself, race a rule-based baseline, or watch a real decision model choose between flap and wait on seeded pipe sequences.
Jev Demos & Experiments#community#github#game
Community
NitroAMD/SystemOne-DemoPython
NitroAMD/SystemOne-DemoPythonA Python dispatcher skeleton for TypeSafe System One (Jev) that turns typed signatures into batched questions, plus a Playwright browser harness that measures the dispatcher against ground truth.
Jev Demos & Experiments#community#github#python
Community
monkey1sai/jev-systemone-labPython
monkey1sai/jev-systemone-labPythonFrom the Jev guides
Original write-ups on what this entry is for.
From the Jev AI model community
Posts from builders shipping with Jev right now.