Online Slip Detection and Friction Coefficient Estimation

Purpose: This repo contains a ROS2 node that detects slip (drift) events and estimates a representative friction coefficient (mu) from IMU + odometry data for experiments used in the associated paper.

Key file:

• src/drift_detector/drift_detector/revised_detector.py — ROS2 Node that:
  • Subscribes to: /ackermann_cmd (AckermannDriveStamped), /odom (Odometry), /sensors/imu/raw (Imu), /odometry/filtered (Odometry)
  • Publishes: Bool on topic is_drifting
  • Computes a slip metric from the difference between filtered and raw odometry magnitudes
  • During detected drift windows computes mu = sqrt(ax^2 + ay^2) / az using IMU linear acceleration
  • Buffers per-event mu values and prints/stores per-run summaries

How detection works:

• If slip_estimate > linear_threshold for longer than drift_length, node enters drifting window
• While drifting, mu samples are appended: mu = sqrt(ax^2 + ay^2) / az
• On the drift end, the node records the maximum mu from the event (used by downstream analysis)

Runtime outputs for parsing:

• The node prints arrays at shutdown:
  • Times: list of drift start timestamps (REMOVE_DRIFT_TIMES)
  • Mus: list of recorded mu values per event (REMOVE_DRIFT_MUS) These prints are intended for downstream parsing and k-fold analysis.

Requirements

• Ubuntu 20.04
• Python 3.8+

  pip install -r requirements.txt

• ROS 2 Foxy
• F1Tenth stack – clone and build the ROS workspace as described in the F1Tenth getting‑started guide:

  # create a drive workspace and clone the system repository
  mkdir -p ~/f1tenth_ws/src
  cd ~/f1tenth_ws/src
  git clone https://github.com/f1tenth/f1tenth_system.git
  # then `colcon build` from the workspace root, source the install/setup.bash, etc.
  

Dataset

• Data/Mu&OffsetData.xlsx - an Excel file with all collected runs including both the offset and mu data.