Back to Projects
Autonomous Systems

Autonomous Decision Engine

View Source Code

Project Overview

Enables spacecraft to autonomously respond to environmental threats and hardware anomalies without waiting for ground intervention, drastically reducing reaction times.

The Problem

The speed of light limits communication with deep space probes. A distress signal from Mars takes up to 22 minutes to reach Earth, and the command to fix it takes another 22 minutes. In critical scenarios, 44 minutes is too long.

Our Solution

An onboard decision engine using reinforcement learning and decision trees. It ingests the output from the anomaly detection models and instantly calculates the optimal sequence of commands to stabilize the spacecraft.

System Architecture

A deterministic state-machine core wrapped in a highly constrained Reinforcement Learning policy network. The engine runs on an isolated, redundant compute node to ensure it remains operational even during primary computer failure.

Impact & Results

Successfully simulated the recovery of a tumbling spacecraft in a frictionless environment in under 3 seconds. Slated for integration into a next-generation asteroid mining probe.

AI Models

Proximal Policy Optimization (PPO)Decision TreesFinite State Machines

Technology Stack

RustC++PythonCore Flight System (cFS)