Skidbot

ROS 2 Jazzy simulation and navigation stack for a custom 4-wheel skid-steer mobile robot, with switchable A* and Hybrid A* global path planners and a full benchmarking pipeline.

Overview

Skidbot is a simulated skid-steer robot built for evaluating kinematically feasible path planning in indoor environments. The robot runs in Gazebo Harmonic and is navigated by the Nav2 stack using either the standard SMAC 2D A* planner or the SMAC Hybrid A* planner (SE(2) state space). A dual Extended Kalman Filter fuses wheel odometry, IMU, and GPS to produce both local (odom) and global (map) pose estimates. Benchmarking nodes record per-run planning metrics, planned path waypoints, and actual robot trajectories to CSV for offline analysis.

This repository accompanies the paper:

K. F. Wassef, "Implementation and Evaluation of Cost-Aware Hybrid-A* Path Planning for a Skid-Steer Mobile Robot in Indoor Environments," Dept. of Mechatronics Engineering, Ain Shams University, 2026.

---

Robot Specifications

Parameter	Value
Chassis (L × W × H)	0.80 m × 0.389 m × 0.171 m
Chassis mass	18.71 kg
Total mass (all links)	~40.0 kg
Wheel radius	0.108 m
Wheel width	0.060 m
Track width (wheel separation)	0.666 m
Wheelbase (front–rear axle)	0.763 m
Drive type	4-wheel skid-steer
Max linear speed	1.0 m/s
Max angular speed	4.0 rad/s

---

Sensor Suite

Sensor	ROS Topic	Rate	Role
2D LiDAR (360°, 12 m)	/scan	—	Obstacle detection, costmap updates
IMU	/imu	100 Hz	Orientation, EKF yaw fusion
Wheel encoders	/skidbot_controller/odom	50 Hz	Odometry, EKF velocity fusion
GPS / NavSat	/navsat	10 Hz	Global position (EKF map-frame anchor)

---

Prerequisites

Requirement	Version
ROS 2	Jazzy
Gazebo	Harmonic (gz-sim 8)

Key ROS packages (install via apt):

sudo apt install \
  ros-jazzy-nav2-bringup \
  ros-jazzy-nav2-smac-planner \
  ros-jazzy-nav2-regulated-pure-pursuit-controller \
  ros-jazzy-robot-localization \
  ros-jazzy-ros-gz-bridge \
  ros-jazzy-ros-gz-sim \
  ros-jazzy-ros2-control \
  ros-jazzy-gz-ros2-control \
  ros-jazzy-diff-drive-controller \
  ros-jazzy-joint-state-broadcaster \
  ros-jazzy-robot-state-publisher \
  ros-jazzy-mapviz \
  ros-jazzy-mapviz-plugins \
  ros-jazzy-swri-transform-util

---

Repository Structure

skidbot-robot/
├── skidbot_description/    # URDF/Xacro robot model, meshes, RViz config
├── skidbot_gazebo/         # SDF simulation worlds, ROS–Gazebo bridge config, world launchers
├── skidbot_controller/     # ros2_control diff-drive controller, twist relay node
├── skidbot_localization/   # Dual EKF (odom + map), NavSat GPS transform, Mapviz launcher
└── skidbot_navigation/     # Nav2 stack, planner configs, behavior trees, benchmarking nodes

---

Build

cd ~/skidbot_ws
colcon build --symlink-install
source install/setup.bash

---

Running

Each step requires its own terminal. Source the workspace in every terminal before running.

Terminal 1 — Gazebo simulation

Choose one environment:

# Open terrain — used for the planner benchmarks in the paper
ros2 launch skidbot_gazebo open_world.launch.py

# Indoor home environment
ros2 launch skidbot_gazebo home_world.launch.py

# Maze environment
ros2 launch skidbot_gazebo maze_world.launch.py

This launches Gazebo, spawns the robot, starts the ROS–Gazebo topic bridge, and opens RViz.

Terminal 2 — Controllers

ros2 launch skidbot_controller controller.launch.py

Starts the diff_drive_controller (ros2_control) and a Twist → TwistStamped relay node.

Terminal 3 — Localization

ros2 launch skidbot_localization localization.launch.py

Starts two EKF nodes:

• ekf_filter_node_odom — fuses odometry + IMU → publishes odometry/local (odom frame)
• ekf_filter_node_map — fuses odometry + IMU + GPS → publishes odometry/global (map frame)

Terminal 4 — Navigation

Choose a planner:

# Hybrid A* (kinematically feasible, SE(2) state space) — recommended for skid-steer
ros2 launch skidbot_navigation nav2.launch.py planner:=hybrid_astar

# Grid A* (SMAC 2D, 8-connected grid)
ros2 launch skidbot_navigation nav2.launch.py planner:=astar

Once Nav2 is active, use the 2D Goal Pose tool in RViz to send navigation goals.

Optional — GPS Mapviz visualization

ros2 launch skidbot_localization mapviz.launch.py

---

Planner Benchmarking

The skidbot_navigation package includes nodes that record per-run planning and navigation metrics.

Record a benchmark run (in a separate terminal, while the robot is navigating):

ros2 run skidbot_navigation planner_benchmark_recorder

Results are saved to skidbot_navigation/planner_results/:

File	Contents
planner_results.csv	Per-run metrics: planning time, path length, heading changes, recoveries, success, nav time
plan_paths.csv	Full planned-path waypoint sequences (x, y, yaw per pose)
robot_trajectories.csv	Actual robot trajectory samples during execution

Open-world results summary

Tests were run in the open-world environment only. Results from the paper:

Planner	Runs	Success	Recoveries	Avg plan time
Hybrid A*	2	2/2 (100%)	0	3.5 ms
A*	4	3/4 (75%)	3 (one failed run)	2.5 ms

The A* failure was caused by the robot facing 180° away from the goal — NavFn ignores initial heading, forcing the Regulated Pure Pursuit controller into repeated recovery maneuvers before aborting.

---

Package Descriptions

skidbot_description

Defines the robot model in URDF/Xacro. Contains the main skidbot.xacro (chassis, arms, wheels, inertia), Gazebo sensor plugins (skidbot.gazebo), and the ros2_control joint interfaces (skidbot_ros2_control.xacro). Meshes are provided as DAE files. The launch file starts robot_state_publisher and optionally opens RViz.

skidbot_gazebo

Three SDF simulation worlds (open, home, maze) with a shared ROS–Gazebo bridge config (skidbot_ros_bridge.yaml) that maps eight topics: /clock, /odom, /joint_states, /tf, /tf_static, /imu, /scan, /navsat. Each world has its own launch file that starts Gazebo, spawns the robot, starts the bridge, and opens RViz.

skidbot_controller

Configures the diff_drive_controller from ros2_control for the 4-wheel skid-steer drive. The controller YAML sets wheel geometry (radius 0.108 m, separation 0.666 m) and velocity/acceleration limits. A twist_to_twist_stamped.py relay node bridges the Nav2 /cmd_vel (Twist) output to the controller's TwistStamped input.

skidbot_localization

Runs two instances of the robot_localization EKF — one anchored to the odom frame (odometry + IMU) and one to the map frame (odometry + IMU + GPS). A navsat_transform_node converts raw GPS fixes to Cartesian odometry. The Mapviz launcher adds a 2D GPS overlay for visualising waypoint routes.

skidbot_navigation

Holds all Nav2 configuration: planner YAMLs for A* and Hybrid A*, Regulated Pure Pursuit controller config, costmap settings, recovery behavior config, and two behavior tree XMLs (one per planner). Python nodes in the package provide GPS waypoint logging, interactive waypoint following from Mapviz clicks, and the planner benchmarking recorders that produce the CSV results files.

---

Key Configuration Files

File	Purpose
skidbot_controller/config/skidbot_controller.yaml	Wheel geometry, velocity/acceleration limits
skidbot_localization/config/dual_ekf_navsat_params.yaml	EKF sensor fusion config for both frames
skidbot_navigation/config/planner_hybrid_astar.yaml	Hybrid A* tuning (turning radius, angle bins, cost penalty)
skidbot_navigation/config/planner_astar.yaml	SMAC 2D A* tuning
skidbot_navigation/config/controller.yaml	Regulated Pure Pursuit controller params
skidbot_gazebo/config/skidbot_ros_bridge.yaml	Gazebo ↔ ROS 2 topic bridge mappings