@@ -10,7 +10,7 @@ configs into Genesis recorders during build, then binds them to the live env aft
 
 ```text
 RecordingCfg(source, field, env_idx)
-└── PyQtPlotCfg / MPLPlotCfg / NPZFileCfg / CSVFileCfg / VideoFileCfg
+└── PyQtPlotCfg / MPLPlotCfg / MPLImagePlotCfg / NPZFileCfg / CSVFileCfg / VideoFileCfg
 ```
 
 ## Sources
@@ -28,12 +28,17 @@ returns a scalar or when a file should capture the full batch.
 
 | Output | Use case |
 |---|---|
-| `PyQtPlotCfg` | Live PyQtGraph plots. |
-| `MPLPlotCfg` | Live Matplotlib plots. |
+| `PyQtPlotCfg` | Live PyQtGraph line plots (time series). |
+| `MPLPlotCfg` | Live Matplotlib line plots (time series). |
+| `MPLImagePlotCfg` | Live Matplotlib image window for a `CameraSensor` source — `field="rgb"` for the frame, `field="depth"` for the depth map. |
 | `NPZFileCfg` | Compressed array dumps at cleanup/reset. |
 | `CSVFileCfg` | Row-oriented streaming output. |
 | `VideoFileCfg` | Camera frames to `.mp4`. |
 
+Line plots (`PyQtPlotCfg` / `MPLPlotCfg`) consume 1-D channel data; `MPLImagePlotCfg` consumes a 2-D
+frame and requires a camera source. All plot windows auto-detect the display and no-op on a headless
+host.
+
 ## Where to continue
 
 - [Sensors](sensors.md)
 
@@ -8,7 +8,7 @@ Recording 把选定运行时信号变成实时曲线、文件或视频，而不
 
 ```text
 RecordingCfg(source, field, env_idx)
-└── PyQtPlotCfg / MPLPlotCfg / NPZFileCfg / CSVFileCfg / VideoFileCfg
+└── PyQtPlotCfg / MPLPlotCfg / MPLImagePlotCfg / NPZFileCfg / CSVFileCfg / VideoFileCfg
 ```
 
 ## 数据源
@@ -25,12 +25,15 @@ RecordingCfg(source, field, env_idx)
 
 | Output | 用途 |
 |---|---|
-| `PyQtPlotCfg` | 实时 PyQtGraph 曲线。 |
-| `MPLPlotCfg` | 实时 Matplotlib 曲线。 |
+| `PyQtPlotCfg` | 实时 PyQtGraph 折线图（时间序列）。 |
+| `MPLPlotCfg` | 实时 Matplotlib 折线图（时间序列）。 |
+| `MPLImagePlotCfg` | `CameraSensor` 源的实时 Matplotlib 图像窗口——`field="rgb"` 显示画面，`field="depth"` 显示深度图。 |
 | `NPZFileCfg` | cleanup/reset 时写压缩数组。 |
 | `CSVFileCfg` | 行式流式输出。 |
 | `VideoFileCfg` | 相机帧写 `.mp4`。 |
 
+折线图（`PyQtPlotCfg` / `MPLPlotCfg`）消费一维通道数据；`MPLImagePlotCfg` 消费二维帧，且要求相机源。所有绘图窗口都会自动检测显示器，无显示器时静默跳过。
+
 ## 继续阅读
 
 - [传感器](sensors.md)
 
@@ -21,7 +21,8 @@ This lazy cache keeps repeated observation and reward reads consistent within on
 | `CameraSensor` | RGB-D camera attached to a link. |
 | `ContactSensor` | Per-link contact force and optional air/contact time. |
 | `FrameTransformerSensor` | Target frame poses relative to a source frame. |
-| `RayCastSensor` | Grid, ring, or hemisphere ray patterns. |
+| `RayCastSensor` | Analytic ray-vs-ground / heightfield (grid, ring, hemisphere patterns) — the fast O(1) path for terrain height-scan. |
+| `MeshRayCastSensor` | Genesis BVH ray-cast against real scene meshes — rays hit arbitrary rigid bodies (props, obstacles). Targets must set `use_visual_raycasting=True`. |
 | `TerrainHeightSensor` | Downward height scan for terrain-aware locomotion. |
 | `SelfContactSensor` | Self-collision/contact signals. |
 | `RootAngularMomentumSensor` | Root angular momentum diagnostics. |
 
@@ -19,7 +19,8 @@ Genesis 不会通过 MJCF sensor block 暴露所有信号，因此 GeneLab 提
 | `CameraSensor` | 挂到 link 上的 RGB-D 相机。 |
 | `ContactSensor` | 每 link 接触力和可选 air/contact time。 |
 | `FrameTransformerSensor` | target frame 相对 source frame 的姿态。 |
-| `RayCastSensor` | grid、ring、hemisphere ray pattern。 |
+| `RayCastSensor` | 解析法射线 vs 地面 / heightfield(grid、ring、hemisphere pattern)——地形高度扫描的 O(1) 快路。 |
+| `MeshRayCastSensor` | Genesis BVH 射线对真实场景 mesh 求交——射线能打到任意刚体(道具、障碍)。目标需设 `use_visual_raycasting=True`。 |
 | `TerrainHeightSensor` | 地形感知 locomotion 的向下 height scan。 |
 | `SelfContactSensor` | 自碰撞/自接触信号。 |
 | `RootAngularMomentumSensor` | root angular momentum 诊断。 |
 
@@ -1,16 +1,18 @@
 # GeneLab visual showcase
 
-Play-only GeneLab extension that wires every M1–M4 building block into a six-task viewer
+Play-only GeneLab extension that wires every M1–M4 building block into an eight-task viewer
 demo. Each task drops a real robot (Franka or Unitree G1) into a `ManagerBasedRlEnv`,
 runs a scripted action loop, and dumps the relevant sensor / curriculum / terrain state
 so the visual behaviour can be checked by eye.
 
-- `GeneLab-Sensors-Showcase-v0` — Franka with `CameraSensor` (RGB+depth → PNG), `IMUSensor`, `FrameTransformerSensor`.
+- `GeneLab-Sensors-Showcase-v0` — Franka with `CameraSensor` (RGB+depth → PNG), `IMUSensor`, `FrameTransformerSensor`, and `ForceTorqueSensor`.
 - `GeneLab-RayCast-Showcase-v0` — Franka with three `RayCastSensor` instances mounted side by side (`GridPattern` + `RingPattern` + `HemispherePattern`), all three dumped in parallel.
 - `GeneLab-Contact-Showcase-v0` — Unitree G1 with `ContactSensor`, air-time tracking on both feet.
 - `GeneLab-Terrain-Showcase-v0` — Unitree G1 dropped on a 1×5 row that tiles every built-in sub-terrain (`flat` + `stairs` + `rough` + `slope` + `wave`) in a single scene.
 - `GeneLab-Curriculum-Showcase-v0` — Unitree G1 on a 5×5 RandomRough grid, scripted XY drift drives `terrain_levels_vel` promotion.
 - `GeneLab-Actuator-Showcase-v0` — Franka with `IdealPDActuator` on the arm (force-channel control), prints joint-tracking error against a sine target.
+- `GeneLab-MlpResidual-Actuator-Showcase-v0` — Franka with `MlpResidualActuator` on the arm (DC-motor base + TorchScript residual network).
+- `GeneLab-Recording-Showcase-v0` — Franka with live PyQt / Matplotlib plots and NPZ / CSV data dumps from an IMU.
 
 ## Installation
 
@@ -55,12 +57,13 @@ examples/genelab_showcase/
 └── src/genelab_showcase/
     ├── tasks.py               # register() + RegisteredTask dispatcher
     ├── runner.py              # ShowcaseRunner base class
-    ├── sensors/               # Camera + IMU + FrameTransformer
+    ├── sensors/               # Camera + IMU + FrameTransformer + ForceTorque
     ├── raycast/               # Grid / Ring / Hemisphere
     ├── contact/               # G1 foot air-time
     ├── terrain/               # 5 sub-terrains + mixed
     ├── curriculum/            # terrain_levels_vel
-    └── actuators/             # ImplicitPD / IdealPD / DCMotor
+    ├── actuators/             # IdealPD (force control) + MlpResidual (DC-motor + residual)
+    └── recording/             # live PyQt/MPL plots + NPZ/CSV IMU dump
 ```
 
 The runtime entry point is `genelab_showcase.tasks:register`, auto-discovered by
 
@@ -0,0 +1,50 @@
+"""Lazy pyqtgraph window helper shared by the showcase runners.
+
+Every showcase that renders a live Qt window (point cloud, camera image, plot) opens it lazily
+on the first viz tick and tolerates a missing pyqtgraph / display: the window is built once via
+a builder callback, and if that fails the runner silently continues without it (never retrying).
+"""
+
+from collections.abc import Callable
+from typing import Any
+
+
+class LazyQtWindows:
+    """One-time pyqtgraph app + window setup with a best-effort fallback.
+
+    Usage::
+
+        self._qt = LazyQtWindows()
+        ...
+        if self._qt.ensure(self._build, label="my showcase"):
+            self._item.setData(...)
+            self._qt.process()
+
+    :meth:`ensure` calls ``builder(app)`` exactly once (``app`` is the ``QApplication``). It
+    returns ``True`` when pyqtgraph is importable and the build succeeds, ``False`` otherwise —
+    and caches the verdict so a failed import is not retried on every step.
+    """
+
+    def __init__(self) -> None:
+        self._app: Any = None
+        self._tried = False
+        self._ready = False
+
+    def ensure(self, builder: Callable[[Any], None], *, label: str) -> bool:
+        if self._tried:
+            return self._ready
+        self._tried = True
+        try:
+            import pyqtgraph as pg
+
+            self._app = pg.mkQApp(label)
+            builder(self._app)
+            self._ready = True
+        except Exception as exc:  # noqa: BLE001 - viz is best-effort, never break the sim
+            print(f"{label}: live window unavailable ({exc})")
+        return self._ready
+
+    def process(self) -> None:
+        """Pump the Qt event loop so the window stays responsive."""
+        if self._app is not None:
+            self._app.processEvents()