Differential Drive Robot — ROS 2 Controllers + Plots

Repository contains a differential-drive robot model for Gazebo (ROS 2 Humble) and four simple path-following controllers:

• P waypoint controller
• Pure Pursuit
• Stanley
• MPC (minimal brute-force grid)

Each controller can be launched together with an odometry plotter that saves a 2×2 figure:

1. XY trajectory (actual vs expected waypoint polyline + tolerance circles)
2. x(t)
3. y(t)
4. heading ϕ(t)

Plots are saved into ./plots/.

---

Repository structure

• my_robot_description/ — URDF/Xacro + meshes + RViz config
• my_robot_bringup/ — Gazebo launch (my_robot_gazebo.launch.xml), keyboard control, odom listener
• my_robot_controllers/ — controllers + plotter + launch files
  • controllers: p_controller.py, pure_pursuit_controller.py, stanley_controller.py, mpc_controller.py
  • plotter: odom_plotter.py
  • launches: *_with_plots.launch.py
• plots/ — saved plot images (example runs)

---

Controllers overview

P (waypoint-to-waypoint)

Moves to the current waypoint using:

• linear speed proportional to distance error (k_rho * rho)
• angular speed proportional to heading error (k_alpha * alpha) Optionally slows down when heading error is large (to avoid “driving sideways”).

Pure Pursuit

Chooses a lookahead point on the path at distance Ld and computes curvature:

• kappa = 2*sin(alpha) / Ld
• w = v * kappa Works well for smooth paths; can cut corners if lookahead is too large.

Stanley

Uses:

• heading error to the path direction
• cross-track error (signed distance to closest point on the path) Steering:
• delta = e_psi + atan2(k * e_cte, v + v0) Then mapped to diff-drive angular velocity via w = (v/L)*tan(delta) (with a clamp on delta).

MPC (minimal)

Very simple MPC-like selection:

• pick a target waypoint (nearest + a few points ahead)
• brute-force a small grid of constant (v,w) commands for horizon N
• simulate kinematic rollout and choose the lowest cost

---

Quick start (Docker)

1) Clone

git clone [email protected]:Artyom35689/AutMobRob-Assignment1.git
cd AutMobRob-Assignment1

2) Build + start container

docker compose build
docker compose up -d

3) Open shell in container

docker compose exec terminal bash

4) Build workspace

cd ~/ros2_ws
colcon build
. install/setup.bash

5) Run controller + plots (one of)

ros2 launch my_robot_controllers p_with_plots.launch.py
ros2 launch my_robot_controllers pure_pursuit_with_plots.launch.py
ros2 launch my_robot_controllers stanley_with_plots.launch.py
ros2 launch my_robot_controllers mpc_with_plots.launch.py

After completion the plot will appear in ./plots/ on the host.

---

Run without Docker (host Ubuntu 22.04 + ROS 2 Humble)

1) Clone and build

git clone [email protected]:Artyom35689/AutMobRob-Assignment1.git
cd AutMobRob-Assignment1
colcon build
. install/setup.bash

2) Launch controller + plots

Same commands as above.

---

Parameters (most useful)

All controllers share:

• odom_topic (default /odom)
• cmd_vel_topic (default /cmd_vel)
• waypoints_flat (default polyline in code)
• v_max, w_max
• control_rate

Plotter:

• plot_name — name used in the figure title and filename prefix
• output_dir — where to save plots (default points to repo plots/)
• goal_tolerance
• stop detection: stop_lin_eps, stop_ang_eps, settle_time
• safety: max_duration

---

Example output plots

See ./plots/:

• run_p_controller_done_*.png
• run_pure_pursuit_done_*.png
• run_stanley_controller_done_*.png
• run_mpc_controller_done_*.png