Video summary

09242021 Lawn Tractor Meeting

Main summary

Key takeaways

Technology

Overview

The meeting focused on integrating ROS with a small lawn-tractor robot. Main topics included path planners, STM32-based controllers, flashing/debugging tools, odometry and steering calibration, and ROS tooling (plotting, topics, and documentation). Community items mentioned included a commercial agricultural route/path planner from an Argentinian developer/company (not free) and recent/planned releases from ROS/ROS-Industrial and local developer communities (Ecumen).

Hardware and flashing / debugging tools

Boards discussed

  • STM32F4 Discovery — includes on-board ST‑LINK debugger.
  • “Blue Pill” STM32 dev board — typically requires an external ST‑LINK or an alternate Arduino-compatible bootloader.

Tools

  • STM32CubeProgrammer (STMicroelectronics)
    • Used to flash .hex files and to upgrade ST‑LINK firmware.
    • Has a Java dependency; behavior varies by OS (Linux/Windows).
  • STM32CubeIDE
    • STM IDE; sometimes fails to communicate with ST‑LINK. CubeProgrammer can be used as a workaround.
  • ST‑LINK debugger/cable
    • Required for instruction-level debugging (breakpoints, single-step, variable/assembly viewing).
  • Alternative: Arduino-compatible bootloader on Blue Pill
    • Flashing the Blue Pill with an Arduino bootloader allows USB programming via the Arduino IDE when ST‑LINK is unavailable.

Practical notes and troubleshooting

  • Discovery board is recognized automatically as ST‑LINK when plugged in; Blue Pill usually needs wiring or an external ST‑LINK.
  • ST‑LINK firmware may need upgrading via the programmer UI.
  • Troubleshooting tips: check device detection, ST‑LINK firmware versions, and OS-specific port permissions (Linux quirks).

Path planners, ROS planners and conversion utilities

Planners referenced

  • TEB (Timed Elastic Band) local planner — teb_local_planner
    • Can be configured to produce steering-angle style outputs (angle instead of ω).
  • Default local planners — trajectory_rollout, DWA
    • Typically output linear and angular velocities (v, ω).
  • Noted: ROS-Industrial and community planner posts exist (some older than initially thought).

Conversion utility

  • A ROS node/utility exists (in teb_local_planner or related repos) to convert standard cmd_vel (v, ω) into Ackermann-style messages (steering angle + forward speed) or other robot-specific command formats.
  • Important implementation note: guard against divide-by-zero when ω or v = 0 — return zero steering angle in that case.

Steering geometry, odometry and calibration

Key kinematic relations

  • Turning radius:
    • R = v / ω (when ω ≠ 0)
  • Ackermann-style steering angle (front wheel):
    • δ = atan(L / R), where L is the wheelbase (distance from front axle to rear axle)

Note: the relation above comes from viewing the turnaround radius and wheelbase; some descriptions use atan(wheelbase / radius).

Odometry from differential wheel encoders

  • Forward distance ≈ (Δleft + Δright) / 2
  • Instantaneous rotation (radians) ≈ (Δleft − Δright) / track_width
    • track_width = distance between rear wheel contact points
  • Heading update: integrate rotation samples over time (heading += Δθ each sample)

Recommended calibration procedure

  1. Physically measure maximum wheel angles at hard stops:
    • Mark tire positions for left hard stop, center, and right hard stop (chalk/wood block).
    • Measure angles with a protractor to get max steering angles.
  2. Record encoder/A-D counts corresponding to left hard stop, center, right hard stop.
  3. Map planner output angles to encoder counts (e.g., input angle range ±maxDegrees → encoder range).
  4. Verify mapping:
    • Command the vehicle to drive at a given angle, read odometry-derived rotation, and compare to expected rotation.
    • Iterate mapping parameters until odometry and steering outputs match.
  5. Ensure accurate physical measurements for wheelbase L and track width (example used in meeting: ~0.51–0.52 m).

Debugging, telemetry and development workflow

Debug data strategy

  • Microcontroller publishes a single comma-separated array (stm_array) containing many floating values (odom results, encoder counts, PWM outputs, steering PWM, instantaneous radius, heading delta, etc.).
    • Advantage: simplifies logging and plotting rather than creating many small topics.
  • A ROS node parses the array and:
    • Builds/publishes standard odometry topics and TF transforms.
    • Optionally extracts and republishes individual values as dedicated topics.
  • Publish rate is controlled by the microcontroller loop (example: 10 Hz); ROS receives and republishes at that rate.

Debugging approaches

  • Arduino-style debugging: typically uses serial print statements.
  • ST‑LINK (GDB-style) debugging: allows stepping through C code, setting breakpoints, watches, and inspecting assembly — more powerful for low-level issues.

Performance note

  • Some Python ROS nodes were observed consuming 100% CPU — likely due to missing proper ROS spin/sleep (use rospy.Rate or similar). Fix spin loops to reduce CPU load.

Plotting and analysis

  • PlotJuggler used for visualizing topics and comparing values; it aligns data by timestamp even if topics publish at different rates.
  • Recent PlotJuggler versions reportedly auto-convert quaternion topics to Euler angles for plotting.
  • Use PlotJuggler to compare odometry, IMU, GPS and microcontroller values by timestamp for validation and tuning.

Guides, tutorials and references

Mentioned or implied resources:

  • STM32CubeProgrammer docs/website (flashing and ST‑LINK firmware updates).
  • YouTube video on flashing Blue Pill with an Arduino bootloader (enables Arduino IDE uploads).
  • teb_local_planner GitHub (conversion node for v/ω → Ackermann/steering).
  • PlotJuggler (visualization and bag file analysis).
  • ROS wiki pages (package docs across kinetic/melodic/noetic — noted as sometimes hard to navigate).

ROS documentation and wiki translation issues

  • ROS1 wiki (MoinMoin) is complex and inconsistent; translating package pages across distributions (kinetic/melodic/noetic) is difficult.
  • ROS2 docs (ReadTheDocs/GitHub) are easier to clone and edit — suggested model for mirroring ROS1 docs.
  • Proposed approach: scrape the ROS wiki to collect pages and analyze/repair broken links offline (direct server access not available). Community translation work is ongoing; several Spanish contributors involved.

Recommendations / Action items

  • For STM32 development:
    • Obtain a working ST‑LINK adapter and use STM32CubeProgrammer to flash and update ST‑LINK firmware.
    • Configure STM32CubeIDE to talk to ST‑LINK for full debugging.
  • If ST‑LINK is not available:
    • Consider flashing a Blue Pill with an Arduino bootloader for rapid iteration via USB.
  • Software and planner integration:
    • Use or modify the teb_local_planner conversion node to accept cmd_vel and output ackermann-style commands; ensure divide-by-zero checks.
  • Calibration:
    • Calibrate steering by measuring physical max angles and mapping encoder counts accordingly; validate against odometry-derived rotation.
  • Performance:
    • Fix Python ROS nodes to use proper rospy spin and Rate controls to avoid 100% CPU usage.
  • Analysis:
    • Use PlotJuggler to compare and validate odometry, IMU, GPS, and microcontroller values by timestamp.

Main speakers / sources mentioned

  • Meeting participants (unnamed): primary developer presenting STM32, odometry, steering calibration, and ROS integration; another participant assisting with ST‑LINK/CubeProgrammer troubleshooting and tips.
  • External organizations / projects:
    • Ecumen (Argentina) — local developers releasing path planner(s) and Arduino/UNO port.
    • Argentinian developer/company — commercial agricultural path planner (cloud farming/route planner; not free).
    • STMicroelectronics — ST‑LINK, STM32CubeProgrammer, STM32CubeIDE.
    • TEB planner / teb_local_planner GitHub and ROS planner ecosystem (move_base, trajectory_rollout, DWA).
    • PlotJuggler (visualization tool).
    • ROS wiki / ROS community contributors (including Matias and other translators).

Original video