appRobotHoming/scripts/5_pose_estimation.py

#!/usr/bin/env python3
"""
pose_estimation.py
==================
Estimate robot joint angles (x, y, z, a, b, c, e) from triangulated marker
poses, using the kinematic model in robot.json (via robot_fk.py).

Design
------
The estimator is parametrised over JOINT VARIABLES, not links. This handles the
tricky cases of this robot family generically:
  * Links with zero own markers (Base/x, Hand/b, Palm/c) — their variable is
    observable only through descendant markers.
  * A variable shared by several links (FingerA & FingerB share 'e').
  * Occluded middle links — global BA reconstructs them from the fingers.

Four switchable methods (robot.json -> pose_estimation.method):
  sequential_vector : analytic per joint from marker-pair / normal vectors (fast)
  sequential_fk     : block-wise least squares along the chain (robust, 1 marker ok)
  global_ba         : all variables jointly, position + normal residuals, robust loss
  hybrid            : sequential_fk init -> global_ba refine   (default, most stable)

Observation input:
  marker_observation = "corner_pose"  -> aruco_marker_poses.json (pos + measured normal)
  marker_observation = "center_point" -> aruco_positions_*.json   (pos only)

Homing integration (appRobotHoming, see doc/Homing_5_Pose.md):
  --from-state <json>        seed/init state (flat {var: value}, or the
                              {"accumulated_state": {...}} shape written by
                              4b_revolute_angle.py) used as x0 for
                              global_ba/hybrid instead of the internal
                              estimate_sequential_fk() cold start. Missing
                              variables default to 0 and are estimated/flagged
                              normally. Without --from-state, behaviour is
                              unchanged (internal cold start, as before).
  robot.json -> pose_estimation.fit_origin_link = "Arm1"   one switch: the
                              named link's jointToParent.origin Y/Z is fit
                              TOGETHER WITH the normal pose in the same
                              global_ba solve (complements the geometric
                              multi-pose method in doc/Kalibrierung.md
                              Schritt [4]) and the result is adopted
                              automatically -- written back into robot.json
                              (surgical text patch, see patch_robot_json_origin()).
                              Off by default (key absent/null).

Unobservable joints (confidence "none") are written as value=null in the
output JSON — never a fabricated 0 (see movements.<var>.observable).

Both the engine (estimate_pose) and a CLI (main) live here.
"""
from __future__ import annotations

import argparse
import json
import math
import os
import re
import sys
import time
from collections import defaultdict
from pathlib import Path
from typing import Any, Dict, List, Optional, Tuple

import numpy as np

sys.path.insert(0, str(Path(__file__).parent))
from robot_fk import RobotFK, STATE_KEYS  # noqa: E402

try:
    from scipy.optimize import least_squares
    HAVE_SCIPY = True
except ImportError:
    HAVE_SCIPY = False


# ==================================================================
# Config
# ==================================================================

DEFAULT_CFG: Dict[str, Any] = {
    "method": "hybrid",
    "marker_observation": "corner_pose",
    "use_normals": True,
    "normal_weight": 100.0,
    "robust_loss": "huber",
    "huber_delta_mm": 8.0,
    "max_iterations": 200,
    "min_cameras_per_marker": 2,
    "finger_block_joints": ["b", "c", "e"],
    "per_link_method": {},
    # One switch: if set to a link name (e.g. "Arm1"), that link's
    # jointToParent.origin Y/Z is fit together with the normal pose (same
    # global_ba solve) and the result is written back into robot.json
    # automatically. None/absent = off, no behaviour change. See
    # doc/Homing_5_Pose.md "Kalibrier-Switch".
    "fit_origin_link": None,
}


def load_pose_cfg(robot_data: Dict[str, Any]) -> Dict[str, Any]:
    cfg = dict(DEFAULT_CFG)
    cfg.update(robot_data.get("pose_estimation", {}) or {})
    return cfg


# ==================================================================
# Observations
# ==================================================================

def load_observations(path: str, use_normals: bool, min_cams: int = 2) -> Dict[int, Dict[str, Any]]:
    """
    Load marker observations. Accepts aruco_marker_poses.json (with measured
    normal + num_cameras) or aruco_positions_*.json (position only).
    Returns: marker_id -> {pos_mm:(3,), normal:(3,)|None, link:str, n_cams:int}
    """
    data = json.load(open(path, "r", encoding="utf-8"))
    out: Dict[int, Dict[str, Any]] = {}
    for m in data.get("markers", []):
        mid = int(m.get("marker_id", m.get("id", -1)))
        if mid < 0:
            continue
        n_cams = int(m.get("num_cameras", 99))
        if n_cams < min_cams:
            continue
        if "position_mm" in m:
            pos = np.array(m["position_mm"], dtype=float)
        elif "position_m" in m:
            pos = np.array(m["position_m"], dtype=float) * 1000.0
        else:
            continue
        nrm = None
        if use_normals and m.get("normal") is not None:
            nv = np.array(m["normal"], dtype=float)
            nn = np.linalg.norm(nv)
            if nn > 1e-9:
                nrm = nv / nn
        out[mid] = {"pos_mm": pos, "normal": nrm, "link": m.get("link", "?"), "n_cams": n_cams}
    return out


def load_seed_state(path: str) -> Dict[str, float]:
    """
    Load a partial/full joint state to use as an optimisation seed (--from-state).

    Accepts either a flat {variable: value} dict, or the
    {"accumulated_state": {...}, ...} wrapper written by 4b_revolute_angle.py —
    same unwrap rule as server/homingOrchestrator.js
    (`stateData.accumulated_state ?? stateData`), so 4b's output files can be
    passed in directly. Unknown keys are ignored; missing STATE_KEYS are simply
    absent from the returned dict (caller defaults them, e.g. to 0.0).
    """
    data = json.load(open(path, "r", encoding="utf-8"))
    raw = data.get("accumulated_state", data) if isinstance(data, dict) else {}
    return {k: float(v) for k, v in raw.items() if k in STATE_KEYS and v is not None}


# ==================================================================
# Kinematic chain analysis
# ==================================================================

def analyze_chain(fk: RobotFK) -> Dict[str, Any]:
    """
    Derive, generically from the FK topology:
      ordered_vars   : movable joint variables, root->tip order, de-duplicated
      var_links      : variable -> list of links it drives
      link_markers   : link -> [model marker ids]
      blocks         : sequential estimation blocks; each block groups the
                       zero-marker ancestor variables with the next marker-
                       bearing joint variable, estimated from that link's own
                       markers (+ siblings sharing the same variable).
    """
    links = fk.links
    topo = fk._topo

    link_markers: Dict[str, List[int]] = {}
    for ln, ld in links.items():
        ids = []
        for mk in ld.get("markers", []) or []:
            if "id" in mk and "position" in mk:
                ids.append(int(mk["id"]))
        link_markers[ln] = ids

    link_var: Dict[str, str] = {}
    for ln, ld in links.items():
        j = ld.get("jointToParent", {}) or {}
        if str(j.get("type", "")).lower() in ("revolute", "linear"):
            v = str(j.get("variable", "")).lower()
            if v:
                link_var[ln] = v

    var_type: Dict[str, str] = {}
    var_links: Dict[str, List[str]] = defaultdict(list)
    for ln, v in link_var.items():
        var_links[v].append(ln)
        var_type[v] = str(links[ln].get("jointToParent", {}).get("type", "")).lower()

    ordered_vars: List[str] = []
    for ln in topo:
        if ln in link_var and link_var[ln] not in ordered_vars:
            ordered_vars.append(link_var[ln])

    # ---- build blocks ----
    blocks: List[Dict[str, Any]] = []
    var_block: Dict[str, int] = {}
    pending: List[str] = []
    for ln in topo:
        if ln not in link_var:
            continue
        v = link_var[ln]
        own = link_markers.get(ln, [])
        if v in var_block:
            # shared variable already in a block -> add this link's markers there
            if own:
                blocks[var_block[v]]["markers"].extend(own)
            continue
        if own:
            bvars = []
            for x in pending + [v]:
                if x not in bvars and x not in var_block:
                    bvars.append(x)
            blocks.append({"vars": bvars, "markers": list(own), "anchor": ln})
            for x in bvars:
                var_block[x] = len(blocks) - 1
            pending = []
        else:
            if v not in pending:
                pending.append(v)
    if pending:
        blocks.append({"vars": pending, "markers": [], "anchor": None})
        for x in pending:
            var_block[x] = len(blocks) - 1

    # subtree_markers[L] = L's own markers + all descendants' markers. Lets
    # observability() credit a block whose own link saw nothing this capture
    # but whose CHILD link did (e.g. Ellbow has no visible markers, but Arm2's
    # markers still constrain z through the chain — same idea as 4b's
    # Fallback-1, just for confidence reporting here, not for the fit itself).
    children: Dict[str, List[str]] = defaultdict(list)
    for ln, ld in links.items():
        p = ld.get("parent")
        if p:
            children[p].append(ln)
    subtree_markers: Dict[str, List[int]] = {}
    for ln in reversed(topo):
        ids = list(link_markers.get(ln, []))
        for c in children.get(ln, []):
            ids.extend(subtree_markers.get(c, []))
        subtree_markers[ln] = ids

    return {
        "ordered_vars": ordered_vars,
        "var_type": var_type,
        "var_links": dict(var_links),
        "link_markers": link_markers,
        "subtree_markers": subtree_markers,
        "blocks": blocks,
    }


# ==================================================================
# Residuals
# ==================================================================

def model_markers(fk: RobotFK, state: Dict[str, float]) -> Dict[int, Dict[str, np.ndarray]]:
    T = fk.compute(state)
    return fk.all_markers_world(T)  # mid -> {world_mm, normal_world, link, local_mm}


def residual_vector(state: Dict[str, float], fk: RobotFK, obs: Dict[int, Dict[str, Any]],
                    marker_ids: List[int], cfg: Dict[str, Any]) -> np.ndarray:
    """Position (mm) + optional normal (scaled) residuals over the given markers."""
    model = model_markers(fk, state)
    res: List[float] = []
    w_n = float(cfg.get("normal_weight", 30.0))
    use_n = bool(cfg.get("use_normals", True))
    for mid in marker_ids:
        if mid not in model or mid not in obs:
            continue
        mm = model[mid]
        dp = np.asarray(mm["world_mm"], float) - obs[mid]["pos_mm"]
        res.extend(dp.tolist())
        if use_n and obs[mid]["normal"] is not None and "normal_world" in mm:
            dn = (np.asarray(mm["normal_world"], float) - obs[mid]["normal"]) * w_n
            res.extend(dn.tolist())
    return np.asarray(res, dtype=float)


def _state_from_vec(var_names: List[str], vec: np.ndarray, base: Dict[str, float]) -> Dict[str, float]:
    s = dict(base)
    for name, val in zip(var_names, vec):
        s[name] = float(val)
    return s


# ==================================================================
# Method: global bundle adjustment
# ==================================================================

def estimate_global_ba(fk: RobotFK, obs: Dict[int, Dict[str, Any]], var_names: List[str],
                       x0: Dict[str, float], cfg: Dict[str, Any]) -> Dict[str, float]:
    if not HAVE_SCIPY:
        print("[WARN] scipy missing — global_ba skipped, returning init")
        return dict(x0)
    marker_ids = list(obs.keys())
    base = {k: 0.0 for k in STATE_KEYS}
    base.update(x0)

    # One switch (pose_estimation.fit_origin_link): also free that link's
    # jointToParent.origin Y/Z as 2 extra parameters of THIS SAME solve, no
    # separate pass. fk.links[...] is mutated in place -- compute() re-reads
    # it fresh every call (robot_fk.py), so this takes effect immediately and
    # is left adopted on success (main() writes it back into robot.json).
    origin_link = cfg.get("fit_origin_link")
    origin = fk.links.get(origin_link, {}).get("jointToParent", {}).get("origin") if origin_link else None
    origin = origin if isinstance(origin, list) and len(origin) == 3 else None
    origin_before = list(origin) if origin else None
    n_state = len(var_names)

    vec0 = np.array([base.get(v, 0.0) for v in var_names]
                     + (origin_before[1:3] if origin else []), dtype=float)

    def fun(vec):
        st = _state_from_vec(var_names, vec[:n_state], base)
        if origin:
            origin[1], origin[2] = float(vec[n_state]), float(vec[n_state + 1])
        return residual_vector(st, fk, obs, marker_ids, cfg)

    loss = cfg.get("robust_loss", "huber")
    f_scale = float(cfg.get("huber_delta_mm", 8.0))
    try:
        sol = least_squares(fun, vec0, loss=loss, f_scale=f_scale,
                            max_nfev=int(cfg.get("max_iterations", 200)) * max(1, len(vec0)))
        state = _state_from_vec(var_names, sol.x[:n_state], base)
        if origin:
            origin[1], origin[2] = float(sol.x[n_state]), float(sol.x[n_state + 1])
            print(f"[INFO] fit_origin_link={origin_link}: Y,Z {origin_before[1:3]} "
                  f"-> [{origin[1]:.3f}, {origin[2]:.3f}]")
        return state
    except Exception as exc:
        print(f"[WARN] global_ba failed: {exc}")
        if origin:
            origin[1], origin[2] = origin_before[1], origin_before[2]  # restore on failure
        return dict(base)


# ==================================================================
# Method: sequential block-wise FK fit
# ==================================================================

def _multistart_values(vtype: str) -> List[float]:
    # revolute: scan the circle to escape local minima at large angles
    if vtype == "revolute":
        return [0.0, 60.0, 120.0, 180.0, 240.0, 300.0]
    return [0.0]


def estimate_sequential_fk(fk: RobotFK, obs: Dict[int, Dict[str, Any]], chain: Dict[str, Any],
                           cfg: Dict[str, Any], seed: Optional[Dict[str, float]] = None
                           ) -> Dict[str, float]:
    """
    Estimate block by block along the chain, freezing already-solved variables.

    seed: optional partial/full state (e.g. from 4b_revolute_angle.py) to trust
    as a starting point. A block is SKIPPED entirely (seed used as-is, no
    re-fit) only if ALL of its variables are present in seed. Blocks with any
    missing variable are still fit normally — including their own multi-start
    — but using the seeded values of EARLIER blocks as fixed context instead
    of 0. This keeps the local-minimum protection for whatever the seed does
    NOT cover (see doc/Homing_5_Pose.md "Wichtige Einschraenkung"), while not
    re-perturbing values the caller already trusts.
    """
    state = {k: 0.0 for k in STATE_KEYS}
    if seed:
        state.update({k: v for k, v in seed.items() if k in STATE_KEYS})
    var_type = chain["var_type"]

    for block in chain["blocks"]:
        bvars = block["vars"]
        bmarkers = [m for m in block["markers"] if m in obs]
        if not bvars:
            continue
        if seed and all(v in seed for v in bvars):
            continue  # fully seeded — trust it, don't re-fit
        if not bmarkers:
            # unobservable block: leave at seed/0, flag later
            continue

        if not HAVE_SCIPY:
            continue

        base = dict(state)

        def fun(vec, _bvars=bvars, _bm=bmarkers, _base=base):
            st = _state_from_vec(_bvars, vec, _base)
            return residual_vector(st, fk, obs, _bm, cfg)

        # multi-start over the first revolute variable in the block
        starts = [[0.0] * len(bvars)]
        lead_type = var_type.get(bvars[0], "linear")
        if lead_type == "revolute":
            starts = []
            for a0 in _multistart_values("revolute"):
                s = [0.0] * len(bvars)
                s[0] = a0
                starts.append(s)

        best, best_cost = None, float("inf")
        for s0 in starts:
            try:
                sol = least_squares(fun, np.array(s0, dtype=float),
                                    loss=cfg.get("robust_loss", "huber"),
                                    f_scale=float(cfg.get("huber_delta_mm", 8.0)),
                                    max_nfev=200 * max(1, len(bvars)))
                if sol.cost < best_cost:
                    best_cost, best = sol.cost, sol.x
            except Exception:
                continue
        if best is not None:
            for name, val in zip(bvars, best):
                state[name] = float(val)

    # wrap revolute angles to (-180, 180]
    for v, vt in var_type.items():
        if vt == "revolute":
            state[v] = (state[v] + 180.0) % 360.0 - 180.0
    return state


# ==================================================================
# Method: sequential analytic vector  (per revolute joint)
# ==================================================================

def estimate_sequential_vector(fk: RobotFK, obs: Dict[int, Dict[str, Any]], chain: Dict[str, Any],
                               cfg: Dict[str, Any]) -> Dict[str, float]:
    """
    Analytic angle from marker geometry where possible. For revolute joints with
    >=2 markers on the link, use the perpendicular marker-pair vector. Falls back
    to the FK block solver for linear / zero-marker / single-marker cases, so it
    always returns a full state (still cheaper than full sequential_fk because
    well-populated joints are solved in closed form).
    """
    state = {k: 0.0 for k in STATE_KEYS}
    var_type = chain["var_type"]
    link_markers = chain["link_markers"]
    var_links = chain["var_links"]

    for block in chain["blocks"]:
        bvars = block["vars"]
        if len(bvars) == 1 and var_type.get(bvars[0]) == "revolute":
            v = bvars[0]
            ln = var_links[v][0]
            mids = [m for m in link_markers.get(ln, []) if m in obs]
            if len(mids) >= 2:
                # model vectors must be expressed in the WORLD frame at angle=0
                # (the link frame is already rotated by the parents y,z,...), so
                # use FK marker world positions with this joint set to 0.
                state_v0 = dict(state)
                state_v0[v] = 0.0
                model_v0 = model_markers(fk, state_v0)
                axis_world = fk.joint_axis_world(ln, state_v0)
                ang = _angle_from_pairs_world(mids, model_v0, obs, axis_world)
                if ang is not None:
                    state[v] = ang
                    continue
        # fallback: block FK fit for this single block
        _fit_single_block(fk, obs, block, var_type, cfg, state)

    for v, vt in var_type.items():
        if vt == "revolute":
            state[v] = (state[v] + 180.0) % 360.0 - 180.0
    return state


def _angle_from_pairs_world(mids: List[int], model_v0: Dict[int, Dict[str, np.ndarray]],
                            obs: Dict[int, Dict[str, Any]], axis_world: np.ndarray) -> Optional[float]:
    from itertools import combinations
    a = np.asarray(axis_world, float)
    a = a / (np.linalg.norm(a) + 1e-12)
    angs, ws = [], []
    for i, j in combinations(mids, 2):
        if i not in model_v0 or j not in model_v0:
            continue
        vm = np.asarray(model_v0[j]["world_mm"], float) - np.asarray(model_v0[i]["world_mm"], float)  # world @ angle 0
        vo = obs[j]["pos_mm"] - obs[i]["pos_mm"]   # observed vector (world, mm)
        vm_p = vm - np.dot(vm, a) * a
        vo_p = vo - np.dot(vo, a) * a
        if np.linalg.norm(vm_p) < 5 or np.linalg.norm(vo_p) < 5:
            continue
        ang = math.atan2(float(np.dot(a, np.cross(vm_p, vo_p))), float(np.dot(vm_p, vo_p)))
        angs.append(ang)
        ws.append(np.linalg.norm(vm_p) * np.linalg.norm(vo_p))
    if not angs:
        return None
    s = sum(w * math.sin(x) for w, x in zip(ws, angs))
    c = sum(w * math.cos(x) for w, x in zip(ws, angs))
    return math.degrees(math.atan2(s, c))


def _fit_single_block(fk, obs, block, var_type, cfg, state):
    if not HAVE_SCIPY:
        return
    bvars = block["vars"]
    bmarkers = [m for m in block["markers"] if m in obs]
    if not bvars or not bmarkers:
        return
    base = dict(state)

    def fun(vec):
        return residual_vector(_state_from_vec(bvars, vec, base), fk, obs, bmarkers, cfg)

    starts = [[0.0] * len(bvars)]
    if var_type.get(bvars[0]) == "revolute":
        starts = [[a0] + [0.0] * (len(bvars) - 1) for a0 in _multistart_values("revolute")]
    best, best_cost = None, float("inf")
    for s0 in starts:
        try:
            sol = least_squares(fun, np.array(s0, float), loss=cfg.get("robust_loss", "huber"),
                                f_scale=float(cfg.get("huber_delta_mm", 8.0)), max_nfev=200 * max(1, len(bvars)))
            if sol.cost < best_cost:
                best_cost, best = sol.cost, sol.x
        except Exception:
            continue
    if best is not None:
        for name, val in zip(bvars, best):
            state[name] = float(val)


# ==================================================================
# Dispatch
# ==================================================================

def observability(chain: Dict[str, Any], obs: Dict[int, Dict[str, Any]]) -> Dict[str, Dict[str, Any]]:
    """
    Per-variable confidence from how well its estimation block is determined.
    A block groups coupled variables (e.g. b,c,e on the fingers); confidence is
    driven by markers-per-variable in that block:
        high   : >= 2 markers per variable (well over-determined)
        medium : >= 1 marker  per variable
        low    : fewer markers than variables (under-determined — distrust!),
                 OR no own markers seen but a child link's markers were
                 (indirect evidence through the chain, e.g. Ellbow via Arm2)
        none   : no markers at all, not even indirectly (variable left at 0)
    """
    info: Dict[str, Dict[str, Any]] = {}
    subtree_markers = chain.get("subtree_markers", {})
    for block in chain["blocks"]:
        seen = [m for m in block["markers"] if m in obs]
        indirect = False
        if not seen and block["anchor"]:
            seen = [m for m in subtree_markers.get(block["anchor"], []) if m in obs]
            indirect = bool(seen)
        nvars = max(1, len(block["vars"]))
        ratio = len(seen) / nvars
        if len(seen) == 0:
            conf = "none"
        elif indirect:
            conf = "low"  # indirect/coupled through a child link, not direct
        elif ratio >= 2.0:
            conf = "high"
        elif ratio >= 1.0:
            conf = "medium"
        else:
            conf = "low"
        for v in block["vars"]:
            info[v] = {"observable": len(seen) > 0, "n_markers": len(seen),
                       "block_vars": len(block["vars"]), "confidence": conf,
                       "block_anchor": block["anchor"], "indirect": indirect}
    return info


def estimate_pose(fk: RobotFK, obs: Dict[int, Dict[str, Any]], cfg: Dict[str, Any],
                  seed: Optional[Dict[str, float]] = None) -> Dict[str, Any]:
    """
    seed: optional partial/full joint state (e.g. from load_seed_state(), the
    4b_revolute_angle.py chain) to trust as a starting point for global_ba/
    hybrid. Passed through to estimate_sequential_fk(), which skips re-fitting
    any block that is FULLY covered by seed and otherwise still applies its
    normal per-block multi-start — so variables the seed does NOT cover keep
    the existing local-minimum protection (see doc/Homing_5_Pose.md "Wichtige
    Einschraenkung") instead of silently defaulting to an unprotected 0.
    sequential_vector ignores seed (no x0 input; left untouched on purpose —
    it is the cheap analytic method, not the one this seeding targets).
    """
    chain = analyze_chain(fk)
    var_names = chain["ordered_vars"]
    method = str(cfg.get("method", "hybrid")).lower()
    obsv = observability(chain, obs)

    if method == "sequential_vector":
        state = estimate_sequential_vector(fk, obs, chain, cfg)
    elif method == "sequential_fk":
        state = estimate_sequential_fk(fk, obs, chain, cfg, seed=seed)
    else:  # global_ba / hybrid (default) — both use the same init->refine path
        init = estimate_sequential_fk(fk, obs, chain, cfg, seed=seed)
        state = estimate_global_ba(fk, obs, var_names, init, cfg)

    # final residual stats over all observed markers
    final_res = residual_vector(state, fk, obs, list(obs.keys()), cfg)
    rms = float(np.sqrt(np.mean(final_res ** 2))) if final_res.size else 0.0

    return {"state": state, "method": method, "observability": obsv,
            "residual_rms": rms, "num_markers": len(obs)}


# ==================================================================
# CLI
# ==================================================================

def patch_robot_json_origin(robot_path: str, link_name: str, yz: Tuple[float, float]) -> bool:
    """
    Surgically rewrite links.<link_name>.jointToParent.origin[1],[2] (Y,Z) in
    the robot.json TEXT in place. robot.json has a hand-curated, compact
    format (markers etc. one per line) -- a full json.load()+json.dump()
    round-trip would reformat the whole file, so this only touches the one
    "origin": [...] array belonging to <link_name> (X left untouched).
    Returns True if a match was found and patched.
    """
    with open(robot_path, "r", encoding="utf-8") as f:
        text = f.read()
    link_m = re.search(r'"%s"\s*:\s*\{' % re.escape(link_name), text)
    if not link_m:
        return False
    origin_m = re.search(r'"origin"\s*:\s*\[([^\]]*)\]', text[link_m.end():])
    if not origin_m:
        return False
    parts = [p.strip() for p in origin_m.group(1).split(",")]
    if len(parts) != 3:
        return False
    parts[1] = f"{float(yz[0]):.4f}".rstrip("0").rstrip(".")
    parts[2] = f"{float(yz[1]):.4f}".rstrip("0").rstrip(".")
    new_array = f"[{parts[0]}, {parts[1]}, {parts[2]}]"
    start = link_m.end() + origin_m.start()
    end = link_m.end() + origin_m.end()
    with open(robot_path, "w", encoding="utf-8") as f:
        f.write(text[:start] + '"origin": ' + new_array + text[end:])
    return True


def main() -> None:
    ap = argparse.ArgumentParser(description="Estimate robot joint angles from marker poses")
    ap.add_argument("markers", help="aruco_marker_poses.json (corner_pose) or aruco_positions_*.json (center)")
    ap.add_argument("-robot", "--robot", required=True)
    ap.add_argument("-out", "--out", default=None)
    ap.add_argument("--method", default=None, help="override robot.json method")
    ap.add_argument("--from-state", default=None, metavar="JSON",
                    help="Seed/init state (flat {var:value} or {accumulated_state:{...}} as "
                         "written by 4b_revolute_angle.py). Used as x0 for global_ba/hybrid "
                         "instead of the internal cold start. See doc/Homing_5_Pose.md.")
    args = ap.parse_args()

    robot_data = json.load(open(args.robot, "r", encoding="utf-8"))
    cfg = load_pose_cfg(robot_data)
    if args.method:
        cfg["method"] = args.method

    fk = RobotFK(robot_data)
    obs = load_observations(args.markers, cfg.get("use_normals", True),
                            int(cfg.get("min_cameras_per_marker", 2)))
    seed = load_seed_state(args.from_state) if args.from_state else None
    print(f"[INFO] method={cfg['method']} | observed markers={len(obs)} | use_normals={cfg.get('use_normals')}"
          + (f" | seed={seed}" if seed else ""))
    if cfg.get("fit_origin_link"):
        print(f"[INFO] fit_origin_link={cfg['fit_origin_link']} -> robot.json wird bei Erfolg aktualisiert")

    result = estimate_pose(fk, obs, cfg, seed=seed)
    st = result["state"]

    print("\nEstimated joint values:")
    for v in ["x", "y", "z", "a", "b", "c", "e"]:
        ob = result["observability"].get(v, {})
        unit = "mm" if v in ("x", "e") else "deg"
        conf = ob.get("confidence", "?")
        tag = "" if ob.get("observable", False) else "  [UNOBSERVABLE -> null]"
        print(f"  {v}: {st.get(v, 0.0):8.2f} {unit}   (markers={ob.get('n_markers','?')}, conf={conf}){tag}")
    print(f"\n[INFO] residual RMS over {result['num_markers']} markers: {result['residual_rms']:.3f}")

    movements = {}
    for v in ["x", "y", "z", "a", "b", "c", "e"]:
        ob = result["observability"].get(v, {})
        observable = ob.get("observable", False)
        movements[v] = {
            # Unobservable -> null, never a fabricated 0 (see module docstring).
            "value": st.get(v, 0.0) if observable else None,
            "unit": "mm" if v in ("x", "e") else "deg",
            "observable": observable,
            "confidence": ob.get("confidence", "none"),
            "n_markers": ob.get("n_markers", 0),
        }
    out = {
        "schema_version": "1.0",
        "created_utc": time.strftime("%Y-%m-%dT%H:%M:%SZ", time.gmtime()),
        "method": result["method"],
        "seeded": seed is not None,
        "movements": movements,
        "residual_rms": result["residual_rms"],
        "num_markers": result["num_markers"],
    }
    out_path = args.out or os.path.join(os.path.dirname(args.markers), "robot_state.json")
    json.dump(out, open(out_path, "w", encoding="utf-8"), indent=2)
    print(f"[INFO] wrote {out_path}")

    # "Uebernehmen": fit_origin_link's Y/Z is already adopted on `fk` (see
    # estimate_global_ba) -- write it back into robot.json itself.
    origin_link = cfg.get("fit_origin_link")
    if origin_link:
        origin = fk.links.get(origin_link, {}).get("jointToParent", {}).get("origin")
        if isinstance(origin, list) and len(origin) == 3:
            if patch_robot_json_origin(args.robot, origin_link, (origin[1], origin[2])):
                print(f"[INFO] robot.json aktualisiert: {origin_link}.jointToParent.origin "
                      f"= [{origin[0]}, {origin[1]:.3f}, {origin[2]:.3f}]")
            else:
                print(f"[WARN] {origin_link}.jointToParent.origin in {args.robot} nicht gefunden — "
                      f"nicht aktualisiert (Wert nur in {out_path} sichtbar)")


if __name__ == "__main__":
    main()