Back
roboticsgamesagents

Jev Autopilot

A Jev autopilot flies a simulated drone across a generated Three.js city and lands on a target pad.

What Jev decides

Throttle, yaw, pitch, roll, whether to commit to landing, and whether to cut the motors.

Code calculates flight geometry and obstacles, Jev chooses four control axes and two landing decisions, and the simulator applies the probability distributions with hard safety limits.

CHOICENOUL
Jev role
Choose bounded flight controls from structured navigation and obstacle state.
Input
Altitude, bearing, speed, target geometry, nearby obstacles, and landing state.
Output
Four Choice distributions and two Noul probabilities.
Decision frequency
Repeated throughout one simulated flight.
Fallback
Hard safety limits remain in code, and manual stick input takes control back immediately.
Author reported · Source reviewed

How it works

Input
Altitude, bearing, speed, target geometry, nearby obstacles, and landing state.
Jev
CHOICE + NOUL
Output
Four Choice distributions and two Noul probabilities.

Reported performance

Reported flight cost

0.01 USD

Fly from pad A through a generated city and land on pad B.

Author reported · Source reviewed

Captured Sep 21, 2026

Overview

Jev Autopilot is a visual demonstration of a typed decision model inside a continuous control loop. The simulator computes altitude, bearing, speed, obstacle clearance, and landing state, then sends a compact situation report to Jev.

One request asks for throttle, yaw, pitch, and roll choices plus Noul decisions for committing to landing and cutting the motors. The repository describes an author-reported cost per completed flight; it should be treated as a demo measurement, not a robotics safety claim.

Language
TypeScript
License
Not declared
Author
Ariel Weinberger
Sources checked Sep 21, 2026

Sources

Related projects