Go2 Speed vs Precision testing by mustafab0 · Pull Request #2138 · dimensionalOS/dimos

mustafab0 · 2026-05-18T03:15:08Z

Problem

Tuning the Go2 base controller was guesswork: the FOPDT plant fit and feedforward gains were hand-vendored, with no single artifact telling an operator "for tolerance X cm, run at speed Y."

Issue: #921

Solution

Two CLI tools producing one versioned config artifact. go2_characterization runs a space-cheap velocity-step system-ID, fits FOPDT per axis (vx/vy/wz), and derives feedforward + a curvature velocity profile; go2_benchmark sweeps the hardcoded baseline controller across a speed ladder and writes back the operating-point map + tolerance→max-safe-speed inversion. One SI harness, sim vs hardware is just which plant the steps drive. Carries only the minimal verified dependency closure from the R&D branch (R&D archived off-repo).

Breaking Changes

None

How to Test

uv run pytest dimos/utils/benchmarking/test_tuning.py -q → 16 pass.
uv run python -m dimos.utils.benchmarking.characterization --mode self-test → artifact, valid_for_tuning=false.
uv run python -m dimos.utils.benchmarking.benchmark --config <self-test artifact> --mode sim --speeds 0.5 → sim pre-check map.
Run dimos run unitree-go2-webrtc-keyboard-teleop to position robot before each test

codecov · 2026-05-18T03:21:37Z

❌ 1 Tests Failed:

Tests completed	Failed	Passed	Skipped
1839	1	1838	28

View the top 1 failed test(s) by shortest run time

dimos.project.test_no_sections::test_no_section_markers

Stack Traces | 0.523s run time

def test_no_section_markers():
        """
        Fail if any file contains section-style comment markers.
    
        If a file is too complicated to be understood without sections, then the
        sections should be files. We don't need "subfiles".
        """
        violations = find_section_markers()
        if violations:
            report_lines = [
                f"Found {len(violations)} section marker(s). "
                "If a file is too complicated to be understood without sections, "
                'then the sections should be files. We don\'t need "subfiles".',
                "",
            ]
            for path, lineno, text in violations:
                report_lines.append(f"  {path}:{lineno}: {text.strip()}")
>           raise AssertionError("\n".join(report_lines))
E           AssertionError: Found 56 section marker(s). If a file is too complicated to be understood without sections, then the sections should be files. We don't need "subfiles".
E           
E             .../drive_trains/fopdt_sim_base/adapter.py:75: # =========================================================================
E             .../drive_trains/fopdt_sim_base/adapter.py:77: # =========================================================================
E             .../drive_trains/fopdt_sim_base/adapter.py:91: # =========================================================================
E             .../drive_trains/fopdt_sim_base/adapter.py:93: # =========================================================================
E             .../drive_trains/fopdt_sim_base/adapter.py:98: # =========================================================================
E             .../drive_trains/fopdt_sim_base/adapter.py:100: # =========================================================================
E             .../drive_trains/fopdt_sim_base/adapter.py:110: # =========================================================================
E             .../drive_trains/fopdt_sim_base/adapter.py:112: # =========================================================================
E             .../drive_trains/fopdt_sim_base/adapter.py:130: # =========================================================================
E             .../drive_trains/fopdt_sim_base/adapter.py:132: # =========================================================================
E             .../drive_trains/fopdt_sim_base/adapter.py:141: # =========================================================================
E             .../drive_trains/fopdt_sim_base/adapter.py:143: # =========================================================================
E             .../control/tasks/velocity_profiler.py:79: # ------------------------------------------------------------------
E             .../control/tasks/velocity_profiler.py:81: # ------------------------------------------------------------------
E             .../control/tasks/baseline_path_follower_task.py:123: # ------------------------------------------------------------------
E             .../control/tasks/baseline_path_follower_task.py:125: # ------------------------------------------------------------------
E             .../control/tasks/baseline_path_follower_task.py:178: # ------------------------------------------------------------------
E             .../control/tasks/baseline_path_follower_task.py:180: # ------------------------------------------------------------------
E             .../control/tasks/baseline_path_follower_task.py:254: # ------------------------------------------------------------------
E             .../control/tasks/baseline_path_follower_task.py:256: # ------------------------------------------------------------------
E             .../utils/characterization/trajectories.py:75: # ---------------------------------------------------------------------------
E             .../utils/characterization/trajectories.py:77: # ---------------------------------------------------------------------------
E             .../utils/characterization/trajectories.py:290: # ---------------------------------------------------------------------------
E             .../utils/characterization/trajectories.py:292: # ---------------------------------------------------------------------------
E             .../utils/characterization/trajectories.py:329: # ---------------------------------------------------------------------------
E             .../utils/characterization/trajectories.py:331: # ---------------------------------------------------------------------------
E             .../utils/characterization/trajectories.py:357: # ---------------------------------------------------------------------------
E             .../utils/benchmarking/test_tuning.py:50: # --- DERIVE ---------------------------------------------------------------
E             .../utils/benchmarking/test_tuning.py:137: # --- artifact round-trip --------------------------------------------------
E             .../utils/benchmarking/test_tuning.py:160: # --- tolerance -> max-safe-speed inversion --------------------------------
E             .../utils/benchmarking/plant.py:120: # --- Vendored fitted FOPDT plant for the Go2 base ------------------------
E             .../utils/benchmarking/plant.py:147: # --- Per-robot profile (single source of truth for robot specifics) -----
E             .../utils/benchmarking/characterization.py:86: # --- self-test (in-process FOPDT plant; NOT robot-valid) -----------------
E             .../utils/benchmarking/characterization.py:152: # --- fit-quality graph (the human-facing deliverable) -------------------
E             .../utils/benchmarking/characterization.py:198: # --- hardware SI (real robot over LCM, operator-gated, safe) -------------
E             .../utils/benchmarking/paths.py:78: # ---------------------------------------------------------------------------
E             .../utils/benchmarking/paths.py:80: # ---------------------------------------------------------------------------
E             .../utils/benchmarking/paths.py:190: # ---------------------------------------------------------------------------
E             .../utils/benchmarking/paths.py:192: # ---------------------------------------------------------------------------
E             .../utils/benchmarking/paths.py:284: # ---------------------------------------------------------------------------
E             .../utils/benchmarking/paths.py:286: # ---------------------------------------------------------------------------
E             .../utils/benchmarking/scoring.py:82: # ---------------------------------------------------------------------------
E             .../utils/benchmarking/scoring.py:84: # ---------------------------------------------------------------------------
E             .../utils/benchmarking/scoring.py:129: # ---------------------------------------------------------------------------
E             .../utils/benchmarking/scoring.py:131: # ---------------------------------------------------------------------------
E             .../utils/benchmarking/scoring.py:199: # ---------------------------------------------------------------------------
E             .../utils/benchmarking/scoring.py:201: # ---------------------------------------------------------------------------
E             .../utils/benchmarking/benchmark.py:107: # --- inlined baseline sim harness (was runner.py + sim_blueprint.py) -----
E             .../utils/benchmarking/benchmark.py:279: # --- hw harness (real robot over LCM; closed-loop baseline) -------------
E             .../utils/benchmarking/benchmark.py:441: # --- benchmark ----------------------------------------------------------
E             .../utils/benchmarking/tuning.py:53: # --- DERIVE tunable constants (documented; single source of truth) -------
E             .../utils/benchmarking/tuning.py:86: # --- Artifact schema -----------------------------------------------------
E             .../utils/benchmarking/tuning.py:214: # --- serialization ---
E             .../utils/benchmarking/tuning.py:264: # --- helpers -------------------------------------------------------------
E             .../utils/benchmarking/tuning.py:313: # --- DERIVE: pure model -> config ---------------------------------------
E             .../utils/benchmarking/tuning.py:398: # --- tolerance -> max-safe-speed inversion (pure) ------------------------

lineno     = 398
path       = '.../utils/benchmarking/tuning.py'
report_lines = ['Found 56 section marker(s). If a file is too complicated to be understood without sections, then the sections should...trains/fopdt_sim_base/adapter.py:93: # =========================================================================', ...]
text       = '# --- tolerance -> max-safe-speed inversion (pure) ------------------------'
violations = [('.../drive_trains/fopdt_sim_base/adapter.py', 75, '    # ================================================...pdt_sim_base/adapter.py', 100, '    # ========================================================================='), ...]

dimos/project/test_no_sections.py:145: AssertionError

To view more test analytics, go to the Test Analytics Dashboard
_{📋 Got 3 mins? Take this short survey to help us improve Test Analytics.}

mustafab0 added 11 commits May 19, 2026 17:43

control tasks for nav2 controller + modifiers

6b96810

plant model for system identification

9cf141f

fopdt modelling and go2 sim adapter

a07300b

dataclassses for plant modelling and tuning

a62226f

system identification harness

80a661b

benchmark tool for go2 trajectory testing

16d0a5f

tests and readme

1c5f9f9

modified keyboard teleop to only publish when keys are pressed

134e80c

updated with hw testing

1dc029f

seperated ff and velocity profiler from benchmark

eb721ad

refactored to make stack robot agnostic

bfe69aa

mustafab0 force-pushed the mustafa/task/go2-controller-tuning branch from 2fb6967 to bfe69aa Compare May 20, 2026 00:43

Provide feedback

Saved searches

Use saved searches to filter your results more quickly

Go2 Speed vs Precision testing#2138

Go2 Speed vs Precision testing#2138
mustafab0 wants to merge 11 commits into
mainfrom
mustafa/task/go2-controller-tuning

mustafab0 commented May 18, 2026 •

edited

Loading

Uh oh!

codecov Bot commented May 18, 2026 •

edited

Loading

Uh oh!

Reviewers

Assignees

Labels

Projects

Milestone

Development

Uh oh!

1 participant

Conversation

mustafab0 commented May 18, 2026 • edited Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

Problem

Solution

Breaking Changes

How to Test

Uh oh!

codecov Bot commented May 18, 2026 • edited Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

❌ 1 Tests Failed:

Uh oh!

Reviewers

Assignees

Labels

Projects

Milestone

Development

Uh oh!

1 participant

mustafab0 commented May 18, 2026 •

edited

Loading

codecov Bot commented May 18, 2026 •

edited

Loading