How mdave9111/cooperative-localisation-under-communication-dropout works
This repository is a local simulation benchmark for comparing dead reckoning, a fixed-gain cooperative observer, and an EKF as communication packet loss varies. The principal workflow is to initialize a rover fleet, advance rover motion, form range packets, apply the selected correction, compute non-anchor RMSE, and plot results. The graph separates simulation orchestration, rover state and motion, and packet-based estimation. The available excerpts show these call paths; unshown details are not inferred.
Main entry points: main.py Monte Carlo runner · agent.py Rover agent · estimators.py Range packet model
Components 10
Simulation benchmark
- Monte Carlo runner
main.py - RMSE evaluation
main.py - Benchmark plot
main.py
Rover state and motion
- Rover agent
agent.py - Kinematics and drift
agent.py - Odometry covariance
agent.py
Cooperative estimation
- Range packet model
estimators.py - Fixed-gain observer
estimators.py - EKF range update
estimators.py
Other components
- Benchmark user
Connections 13
- Benchmark user → Monte Carlo runner: runs benchmark
readme.md - Monte Carlo runner → Rover agent: initializes fleet
main.py - Monte Carlo runner → Kinematics and drift: ticks agents
main.py - Rover agent → Kinematics and drift: updates motion
agent.py - Kinematics and drift → Odometry covariance: propagates uncertainty
agent.py - Monte Carlo runner → Range packet model: creates packets
main.py - Range packet model → Rover agent: reads coordinates
estimators.py - Monte Carlo runner → Fixed-gain observer: selects correction
main.py - Rover agent → Fixed-gain observer: delegates correction
agent.py - Monte Carlo runner → EKF range update: applies EKF update
main.py - Monte Carlo runner → RMSE evaluation: computes error
main.py - Rover agent → RMSE evaluation: provides histories
main.py - Monte Carlo runner → Benchmark plot: plots sweep
main.py