Building a Skid-Steer Robot Without Wheel Odometry

A 40 kg autonomous robot with no wheel odometry, reaching over 30 km/h. Skid-steer robots slip. The faster they go, the worse it gets. Instead of trying to correct for that slip, SSR discards wheel odometry altogether and relies on 3D LiDAR and inertial sensing for localization. This video documents the complete development journey — hardware, software, failures, trade-offs, and the decisions that shaped the final machine. ▶ Code, 3D models & configs (MIT License): https://github.com/nstb-robotics/skid... Explore the full development history in the repository. ────────────────────────────── CHAPTERS 0:00 Opening 1:52 System Overview 3:11 Chapter 1 — Physical Structure 7:10 Chapter 2 — Motor Control 11:43 Chapter 3 — Self-Localization 17:17 Chapter 4 — Path Tracking 19:32 Chapter 5 — System Operations 23:34 Chapter 6 — Presence 31:52 Closing ────────────────────────────── SYSTEM COMPONENTS Hardware 4× off-road hoverboard hub motors (no encoders, Hall sensors only) 2× Flipsky Dual FSESC 4.20 over CAN bus Livox Mid-360 (3D LiDAR + IMU), OAK-D S2 (person detection) GEEKOM GT1 Mega (Ubuntu 22.04, ROS 2 Humble) Software FAST-LIO2 for LiDAR-inertial odometry (no wheel odometry) Nav2 with a custom Dynamic Window Pure Pursuit patch on RPP AVAS real-time audio synthesis, WS2812B LED feedback, DualSense teleop ────────────────────────────── CREDITS FAST-LIO2 — MARS Lab, University of Hong Kong Nav2 / Regulated Pure Pursuit — Shrijit Singh & Steve Macenski (Samsung Research America) Dynamic Window Pure Pursuit — Fumiya Ohnishi (Decwest), Keio University VESC — Benjamin Vedder fl_nav2_helper — attraclab & rasheed ────────────────────────────── This channel usually covers music made with stepper motors and solenoids — SSR is where that "how to control a motor" obsession turned into a 40 kg autonomous robot. #ROS2 #Robotics #AutonomousMobileRobot #SLAM #OpenSource #NVIDIAIsaacSim