---
title: 'RoboAtlas: Surgical Geometry & Active SLAM'
url: https://www.emergentmind.com/topics/roboatlas
type: topic
---

# RoboAtlas: Surgical Geometry & Active SLAM

In the supplied arXiv literature, **RoboAtlas** denotes two distinct constructs. In otolaryngologic robotics, it refers to a geometric atlas of the middle ear and paranasal sinuses derived from CT measurements and expressed as mean dimensions, standard deviations, and 90% ranges for robotic workspace design [2110.01246]. In embodied AI, it denotes a contextual Active SLAM framework that combines OpenRoboVox, frontier exploration, global semantic-map reasoning, and egocentric VLM reasoning through a contextual multi-armed bandit, with evaluation in simulation and on a Unitree Go2 robot [2606.26046]. The shared label therefore spans two different technical programs: patient-specific workspace characterization for endoscopic robot design, and large-scale semantic navigation grounded in 3D mapping.

## 1. Distinct uses of the term in the literature

| Usage | Core definition | arXiv id |
|---|---|---|
| Geometric RoboAtlas | Parametric description of linear dimensions, volumes, and angular limits for middle-ear and sinus workspaces | [2110.01246] |
| Active-SLAM RoboAtlas | Contextual Active SLAM framework using OpenRoboVox and a contextual multi-armed bandit | [2606.26046] |

The distinction is substantive rather than nominal. The 2021 surgical work is an anatomical and design-oriented atlas: it defines workspace bounds for endoscopic robotic applications and explicitly does **not** construct a formal PCA-based Statistical Shape Model [2110.01246]. The 2026 system is a robot autonomy stack: it treats RoboAtlas as a decision-making framework that balances geometric exploration and semantic reasoning in large-scale environments [2606.26046].

A common misconception is to treat both as instances of the same software or robotic platform. The available literature does not support that reading. One use is anatomical and preoperative-design oriented; the other is algorithmic and online-navigation oriented.

## 2. Geometric RoboAtlas for otolaryngologic robotic applications

The geometric RoboAtlas was introduced to help define the workspace for robots studied in otolaryngologic surgery, with the stated aim of helping design and optimize such robots while accounting for patient variability [2110.01246]. The study used scans of several patients of different ages and sexes to determine average workspace size in two anatomically linked endoscopic domains: the middle ear and the paranasal sinuses.

The patient cohorts were modest but explicitly specified. The ear workspace cohort comprised **36 patients (17 right, 19 left), age 2–81 (mean 39 y), sex ratio 1:1**. The sinus workspace cohort comprised **23 patients, age 11–95**, with only one patient younger than 20 years. Exclusions included **congenital malformations (e.g. aplasia)** and **prior surgery with major bone distortion**. All measurements were performed on routine CT acquired in 2018 at Nantes University Hospital, using Vue PACS v11.3 (Carestream) [2110.01246].

Methodologically, the atlas is based on direct PACS measurements rather than volumetric registration. The paper reports that **no volumetric mesh or point-cloud registration was performed** and that **no explicit coordinate system definition or Procrustes-style registration was reported**. Ear landmarks included **CAEd_lateral**, **CAEd_medial**, **CAElength**, **OMheight**, **OMwidth**, and **OMap_length**. Sinus landmarks included **Workspace_depth**, **Workspace_width**, **Nasal_fossae_width**, **Workspace_height**, and **Piriforme_orifice_height**. For each measurement, the study computed **mean $\mu$ and standard deviation $\sigma$** in Excel v.14.5, assessed normality by **$r^2$ of normal-fit curves** with **$r^2 > 0.9$ for ear** and **$> 0.8$ for sinuses**, and extracted **90% percentile ranges** [2110.01246].

The paper is explicit about what RoboAtlas is not. It states that it **does not construct a formal PCA-based Statistical Shape Model**, and that **no point-cloud alignment, covariance matrices or principal modes $\Phi$ were computed, nor shape-parameter vectors $b$ derived**. In that sense, the atlas is a parametric anatomical summary rather than a generative shape model [2110.01246].

## 3. Workspace metrics, design constraints, and limitations of the geometric atlas

The atlas reports linear dimensions, volume proxies, and angular limits for two workspaces: the external auditory canal plus tympanic membrane region, and the paranasal sinus corridors [2110.01246].

| Region | Measurement | Reported value |
|---|---|---|
| External auditory canal | CAEd_lateral | mean 6.1 mm, range 3.9–7.3 mm, $\sigma \approx 0.8$ mm |
| External auditory canal | CAEd_medial | mean 7.9 mm, range 6.2–11.1 mm, $\sigma \approx 1.1$ mm |
| External auditory canal | CAElength | mean 26.9 mm, range 22.5–35.3 mm, $\sigma \approx 3.2$ mm |
| External auditory canal | Canal volume | mean 1.32 cm³ |
| Tympanic membrane workspace | OMheight | 16.2 mm, range 14.1–19.4 mm, $\sigma \approx 1.3$ mm |
| Tympanic membrane workspace | OMwidth | 10.8 mm, range 7.6–12.3 mm, $\sigma \approx 1.3$ mm |
| Tympanic membrane workspace | OMdepth | 5.7 mm, range 3.1–7.2 mm, $\sigma \approx 1.0$ mm |
| Tympanic membrane workspace | Eardrum volume proxy | 0.99 cm³ |
| Paranasal sinuses | Workspace_depth | 77.04 mm, range 59–94 mm, $\sigma \approx 8.5$ mm |
| Paranasal sinuses | Workspace_width | 39.26 mm, range 27–47 mm, $\sigma \approx 5.0$ mm |
| Paranasal sinuses | Nasal_fossae_width | 13.74 mm, range 9–18 mm, $\sigma \approx 2.8$ mm |
| Paranasal sinuses | Workspace_height | 55.39 mm, range 42–67 mm, $\sigma \approx 6.2$ mm |
| Paranasal sinuses | Piriforme_orifice_height | 29.57 mm, range 21–36 mm, $\sigma \approx 3.8$ mm |

The 90% population bounds are central to the atlas’s engineering interpretation. For the ear, **90% of patients** lay within **CAEd_lateral $\in [4.35, 7.20]$ mm**, **CAEd_medial $\in [6.2, 10.2]$ mm**, and **CAElength $\in [22.5, 34.6]$ mm**. For the tympanic membrane workspace, the corresponding 90% ranges were **height $\in [14.3, 19.1]$ mm**, **width $\in [8.2, 13.5]$ mm**, and **depth $\in [4.3, 7.3]$ mm**. For the sinuses, the paper reports **Workspace_depth $\in [63.4, 86.8]$ mm**, **Nasal_fossae_width $\in [10.1, 17.5]$ mm**, **Workspace_height $\in [47.1, 66.9]$ mm**, and **Piriforme_orifice_height $\in [24.2, 34.9]$ mm** for 90% of patients [2110.01246].

The atlas also includes entry-vector constraints. In the ear, the **ideal remote center** is described as being **at canal entrance & tympanic membrane**, allowing **$\pm 45^\circ$ of insertion angles in optimal head position**, with **up to $90^\circ$ in extremes**. In the sinuses, the pivot is placed **at piriform aperture**, with **full max travel of endoscope pivot $\approx 90^\circ$**, and **clinically $\approx 70^\circ$ endoscopes ($\pm 35^\circ$)** [2110.01246].

These measurements are translated into robot-design guidelines. The paper recommends **at least 35 mm of axial travel** for the ear and **up to 95 mm insertion depth** for the sinuses. Angularly, it recommends **$\pm 45^\circ$ about the ear entry point** to cover head-tilting extremes and **$\pm 35^\circ$ pivot for a $70^\circ$ endoscope**, with **full mechanical allowance of $\pm 45$–$50^\circ$** in the sinus case. The reported smallest ear canal diameter of **3.9 mm** leads to the requirement that **robot/tool cross-section must be $\leq 2.7$ mm (endoscope) + auxiliary tools ($\leq 1.4$ mm cannula) with safety margin**. The paper further specifies **endoscope $\phi = 2.7$ mm**, **suction $\phi = 0.8$–$1.4$ mm**, **micro-instruments $\phi = 1.0$–$2.5$ mm**, and recommends designing kinematic chains around these diameters with **at least 0.5 mm clearance** [2110.01246].

Several limitations are explicit. The study reports **sample sizes modest (36 ears, 23 sinuses)**, **under-representation of pediatric (< 10 y) and extreme morphologies**, **no statistically significant differences by side or gender in most measures**, and **slight age effects on canal diameters**. Recommended future work includes increasing the cohort to **> 50 per region**, including **younger children (< 10 y)**, constructing **a true PCA-based statistical shape model of ear canal and sinus cavity surfaces**, and extending the atlas to **other endoscopic sites (e.g. larynx, skull-base cavities)** [2110.01246].

## 4. RoboAtlas as contextual Active SLAM

The 2026 RoboAtlas framework addresses contextual Active SLAM by adaptively balancing geometric exploration and semantic reasoning using **OpenRoboVox**, described as a scalable 3D semantic mapping system [2606.26046]. Its architecture couples metric reconstruction, instance-level semantic fusion, compact scene abstraction, and expert-based decision-making.

The map representation has both geometric and semantic components. Geometrically, OpenRoboVox maintains a **Truncated Signed Distance Field (TSDF) running at ~15 Hz** and an **octree/OctoMap projected into a 2D occupancy grid $M_t^{grid}$ for navigation**. Semantically, it uses **a probabilistic voxel grid** in which each voxel $v$ maintains **a Dirichlet posterior $\theta_v$** over a growing set of global instance IDs $\Gamma_t$. Instance labels are produced per frame by **YOLO-World** for bounding boxes and the **Tokenize Anything model (TAP)** for segmentation masks; the resulting 2D detections are back-projected into the voxel grid and fused via Bayesian updates [2606.26046].

To avoid repeated full-map scans, a background thread maintains a compact **Scene-Dictionary $S_t$** of object instances. On each cycle, it selects a small batch $\Omega_t \subset \Gamma_t$ via a priority union of **temporal recency $P_{rec}$**, **active observation $P_{vis}$**, and **rotational consistency $P_{rot}$**. The dictionary is updated incrementally as
$$
S_{t+1} = S_t \cup \{\mathrm{Extract}(O_i) \mid O_i \in \Omega_t\},
$$
where $\mathrm{Extract}(\cdot)$ computes an instance’s **centroid**, **bounding-box**, and **occupancy probability** from its voxels. The system also includes a **2D Pillar Map**, which projects the 3D voxel set $V$ onto a 2D grid $G$ at resolution $r$, storing in each cell the **top-K MAP-assigned instances ranked by voxel count $\omega_\gamma(u,v)$** [2606.26046].

The implementation emphasizes scalability. Volumetric map expansion is performed in **small cubic blocks $b_k$** to bound peak VRAM,
$$
V_{D'} = \bigcup_{k=1}^{N_b} T(V_D \cap b_k),
$$
and the **geometric (15 Hz)** and **semantic (≈5 Hz)** pipelines run asynchronously on separate threads, with fine-grained locks only at fusion points [2606.26046].

## 5. Contextual multi-armed bandit, semantic reasoning, and VLM experts

RoboAtlas formalizes decision-making as a **contextual multi-armed bandit (CMAB)** with three experts, or arms: **frontier exploration $E_{Frontier}$**, **semantic-map reasoning $E_{SemanticMap}$**, and **egocentric VLM reasoning $E_{EgoVLM}$** [2606.26046].

The context vector is
$$
c_t = [\, m_t^{occ},\, \dot{m}_t^{occ},\, B_t,\, a_{t-1},\, V_t^{vlm},\, V_t^{sim}\,],
$$
where $m_t^{occ}$ is the fraction of grid covered, $\dot{m}_t^{occ}$ is the change in coverage since the last step, $B_t$ is a backtracking indicator, $a_{t-1}$ is the previous arm, $V_t^{vlm}$ is a confidence score from the egocentric VLM, and $V_t^{sim}$ is a soft count of candidate observations above cosine-similarity thresholds $\tau_1 < \tau_2 < \tau_3$,
$$
V_t^{sim} = \tau_1 N_1(t) + \tau_2 N_2(t) + \tau_3 N_3(t).
$$
The bandit uses **LinUCB (Disjoint UCB)**. For each arm $a$, the framework maintains $A_a \in \mathbb{R}^{d \times d}$ and $b_a \in \mathbb{R}^d$, computes
$$
\hat{\theta}_a = A_a^{-1} b_a,
\qquad
p_{t,a} = x_{t,a}^\top \hat{\theta}_a + \alpha \sqrt{x_{t,a}^\top A_a^{-1} x_{t,a}},
$$
selects $a_t = \arg\max_{a \in A_t} p_{t,a}$, and updates $A_{a_t}$ and $b_{a_t}$ using the observed reward [2606.26046].

The reward is defined as
$$
r_t = w_1 \dot{m}_t^{occ} - w_2 B_t + w_3 V_t^{vlm} + w_4 V_t^{sim} + w_{succ}\mathbf{1}_{task\ finished},
$$
with **$w_1 = 1.0$**, **$w_2 = 1.0$**, **$w_3 = 1.0$**, **$w_4 = 1.0$**, and **$w_{succ} = 10.0$**. This formulation explicitly combines geometric exploration signals, anti-backtracking pressure, VLM confidence, semantic similarity, and terminal success [2606.26046].

Each expert has a distinct operational role. The **frontier expert** clusters frontier cells into representative centers $C_t$, computes **information gain $I_t(c)=|U_t \cap S(c)|$**, evaluates path length $\ell_t(c)$ with A*, checks energy feasibility, and selects
$$
g_t = \arg\min_{c \in C_t \ feasible} [-I_t(c) + \alpha_f \ell_t(c)].
$$
The **semantic-map expert** filters the Scene-Dictionary using precomputed **2D potential fields $T_{sim}$** and **neighbor similarity $N_{sim}$**, serializes top-relevant instances into a token-budgeted prompt containing each instance’s **caption**, **centroid $p_i$**, **bounding box $b_i$**, and similarity vector $\sigma_i = [\sigma_i^{targ}, \sigma_i^{neigh}]$, then queries **GPT-4o or Qwen2.5-VL** for a JSON ranking. The **egocentric VLM expert** takes current RGB and depth with spatial annotations on a coarse image grid, queries **GPT-4o or Qwen2.5-VL**, and returns both per-annotation confidence $\phi_{VLM}(O_i)$ and a discrete action in **\{“go-to”, “turn-left”, “turn-right”, “about-face”\}**, rolling out up to **$H=3$ steps** before resubmitting when a turn action is selected [2606.26046].

A second common misconception is that the system merely substitutes a large language or vision-language model for classical exploration. The reported architecture is instead explicitly hybrid: the VLM-based experts are grounded in a large-scale semantic map, and semantic as well as VLM inference are asynchronous so that neither blocks the real-time SLAM thread [2606.26046].

## 6. Empirical performance, ablations, and stated limitations

RoboAtlas was evaluated in **Isaac Sim**, **Habitat (HM3D Val-Unseen, 36 scenes, 278 subtasks)**, and on **real hardware: Unitree Go2 quadruped + Jetson Orin + WiFi→desktop RTX 4090** [2606.26046]. The OpenRoboVox scale test on hardware mapped **1 803 m² over two floors** and recorded **29 588 unique instances in one run**.

| Evaluation | Configuration | Reported result |
|---|---|---|
| Isaac Sim CMAB ablation | Frontier only | SR = 53.3%, SPL = 0.386 |
| Isaac Sim CMAB ablation | Ego-VLM only | SR = 66.7%, SPL = 0.510 |
| Isaac Sim CMAB ablation | Semantic-map only (cold) | SR = 0% |
| Isaac Sim CMAB ablation | RoboAtlas (CMAB) | SR = 100%, SPL = 0.920, average time 780 s |
| Hardware directives | 6 varied language tasks | 100% success, mean standoff 0.22 m, $\sigma = 0.16$ m |
| GOAT-Bench Val-Unseen | Prior SOTA (HIMM with GPT-4o) | SR = 72.8%, SPL = 56.1% |
| GOAT-Bench Val-Unseen | RoboAtlas + GPT-4o | SR = 90.6%, SPL = 53.4% |
| GOAT-Bench Val-Unseen | RoboAtlas + Qwen2.5-VL | SR = 88.8%, SPL = 53.1% |

On **GOAT-Bench “Val Unseen”**, RoboAtlas reports the **highest reported success rate (SR) of 90.6% using GPT-4o**, improving over the strongest prior baseline by **17.8 percentage points in SR**. With **Qwen2.5-VL-7B**, it reports **88.8% SR**, which the paper states **outperforms all baselines using GPT-4o in SR**. Per modality with GPT-4o, the reported SR values are **94.9% for object queries**, **84.6% for descriptions**, and **92.0% for image-based queries** [2606.26046].

Bandit telemetry on GOAT-Bench is also reported: across **2 005 decisions**, the picks were **SemanticMap 38.7% (46.8% of path length)**, **Frontier 16.4% (22.6%)**, and **EgoVLM 45.0% (30.6%)**. The paper interprets this as **stage-dependent allocation: early frontier, then VLM, then semantic exploitation** [2606.26046].

The stated limitations are concrete. **Semantic mapping errors**, including missed detections or association noise, can propagate into planning. **Reliance on remote API calls (GPT-4o)** introduces latency and requires reliable network connectivity. The **reward function is hand-tuned**, and the paper suggests that a learned or more principled formulation could improve generalization. It also notes a **slight SPL penalty in long episodes**, describing this as an Active SLAM trade-off that could be tuned per task [2606.26046].

## 7. Relation to Atlas humanoid research and broader methodological context

RoboAtlas should be distinguished from a broader body of work on the **Boston Dynamics Atlas** humanoid, although there are methodological continuities. Atlas research has addressed **partial footholds and line contacts** through foothold exploration, CoP shifting, polygon cropping, and a **whole-body momentum-based control algorithm** running at **1 kHz**, with **continuous sequences of 4–6 steps on straight line-field in real tests** and **CoP tracking error ≈ 2 cm RMS** [1607.08089]. It has also addressed **large step-ups** through a nonlinear trajectory optimization based on simplified centroidal dynamics, solved with **IPOPT**, reporting that **peak torque dropped to ~96 Nm on the 0.35 m step (a 20% reduction)** and from **~130 Nm down to ~105 Nm (19% reduction)** on a **0.40 m** step [2004.12083].

A separate Atlas navigation line introduced a two-stage rough-terrain planner using a **2.5D height map**, **A*** with a traversability-aware edge cost,
$$
g_{ij} = \|x_j - x_i\| + w_f(1-t_f) + w_s(1-t_s) + w_c c_c,
$$
followed by gradient-descent path smoothing, and demonstrated the integrated pipeline on a **DRC Boston Dynamics Atlas robot** with **9/10 trials on cinder blocks** and **8/10 on stones** [2203.00602]. This suggests a broader methodological continuity in Atlas-related robotics: explicit environment representation, optimization or search over physically meaningful costs, and close integration with whole-body balance control. That continuity is suggestive rather than nominal, because the 2026 RoboAtlas framework was evaluated on a **Unitree Go2 quadruped**, not on Atlas [2606.26046].

Viewed across these works, the name RoboAtlas marks two unrelated but technically rigorous uses. In one, it is a patient-variability-aware anatomical parameterization for designing endoscopic surgical robots [2110.01246]. In the other, it is a contextual Active SLAM architecture in which large-scale 3D semantic maps ground frontier exploration, LLM-based global reasoning, and egocentric VLM guidance [2606.26046]. The coexistence of these meanings is unusual, but the underlying ambition is comparable: to replace underspecified operating assumptions with explicit, quantitatively constrained representations of the workspace.

Source: https://www.emergentmind.com/topics/roboatlas