pith. sign in

arxiv: 2606.06762 · v1 · pith:BTP5B45Enew · submitted 2026-06-04 · 💻 cs.RO

Multi-Robot Planning and Control from CCTV Camera Networks in a Real Warehouse

Luke Robinson , Benjamin Ramtoula , Anas Izaaryene , Paul Newman , Daniele De Martini This is my paper

Pith reviewed 2026-06-28 00:41 UTC · model grok-4.3

classification 💻 cs.RO
keywords multi-robot planningCCTV camera networkswarehouse automationimage-space controloff-board computeun-calibrated camerasfield demonstrationcoordinated fleets
0
0 comments X

The pith

External CCTV networks can coordinate multiple warehouse robots using only off-board compute and image-space planning over an uncalibrated camera graph.

A machine-rendered reading of the paper's core claim, the machinery that carries it, and where it could break.

The paper shows that fleets of mobile robots can be planned and controlled in a real warehouse using nothing but a distributed network of CCTV cameras and edge compute, with all sensing and decision-making moved off the robots themselves. It operates entirely in image space by building a pixel-wise topological graph across camera views and uses a hierarchical planner to assign camera sequences to each robot while planning motions within each view. Coordination happens through a prioritised-then-joint strategy that treats overlapping camera regions as shared resources assigned to only one robot at a time. The approach was tested with four robots and thirty cameras covering six 27-metre aisles, producing mission times and coordination statistics without collisions or deadlocks. The authors present this as the first field demonstration of such multi-robot coordination relying solely on external cameras and off-board resources.

Core claim

A hierarchical planner selects a camera sequence per robot and plans its image-space motion through each view, coordinating the fleet with a prioritised-then-joint strategy that treats overlapping camera regions as shared resources held by one robot at a time to prevent collisions and deadlocks. The system runs entirely over an uncalibrated pixel-wise topological camera graph and was validated in a real warehouse with four robots and 30 cameras across six 27 m aisles.

What carries the argument

The hierarchical planner that selects camera sequences per robot and plans image-space motion through each view, coordinated with a prioritised-then-joint strategy treating overlapping camera regions as shared resources.

If this is right

  • Robots can complete warehouse missions without carrying any task-specific navigation hardware.
  • Wide-area multi-robot operation becomes possible with flexible, uncalibrated camera placement.
  • Coordination statistics and mission times can be achieved by treating camera overlaps as exclusive shared resources.
  • Sensing and compute can be moved entirely off the robots while still preventing collisions and deadlocks.

Where Pith is reading between the lines

These are editorial extensions of the paper, not claims the author makes directly.

  • Centralising all perception and planning could lower per-robot hardware costs in structured indoor settings.
  • The same image-space approach might extend to other fixed-camera environments such as factories or loading bays.
  • Removing the need for onboard sensors could simplify robot maintenance and allow cheaper, simpler platforms.
  • The uncalibrated graph suggests the method tolerates occasional camera movement or addition without full recalibration.

Load-bearing premise

The uncalibrated pixel-wise topological camera graph supplies enough connectivity and overlap for the prioritised-then-joint strategy to prevent collisions and deadlocks across the full warehouse layout.

What would settle it

A warehouse test in which two robots simultaneously occupy an overlapping camera region or reach a deadlock while following the generated plans would show the claim does not hold.

Figures

Figures reproduced from arXiv: 2606.06762 by Anas Izaaryene, Benjamin Ramtoula, Daniele De Martini, Luke Robinson, Paul Newman.

Figure 1
Figure 1. Figure 1: Our setup for deploying multiple robots in a large environment fully covered by CCTV cameras with overlapping viewpoints (coloured regions). Off-boarding all robot compute and planning lets us deploy simple and cheap blind robots that only follow received control commands, making deployments more scalable, robust, and practical. Unlike previous works, we support multiple robots through joint planning and c… view at source ↗
Figure 2
Figure 2. Figure 2: We validate our system in a large-scale deployment in a warehouse. We rely on 30 CCTV cameras to cover the six 27 m warehouse aisles and deploy simultaneously four different robots. 2 Background and Related Work 2.1 Infrastructure-Based Visual Servoing for Mobile Robots Within robot control, the most relevant area is eye-to-hand visual servoing, where fixed external cameras observe the robot and supply the… view at source ↗
Figure 3
Figure 3. Figure 3: The proposed off-board multi-robot control architecture. Robots with no on￾board navigational compute receive control commands over WiFi from an edge server, which processes live video from an uncalibrated CCTV network, performs robot de￾tection and state estimation, and coordinates safe navigation through a hierarchical planner operating on an image-space topological camera graph. High-Level Planning The … view at source ↗
Figure 4
Figure 4. Figure 4: Layout of the mid-sized warehouse used for evaluation: six parallel aisles moni￾tored by 30 overhead RGB cameras that fully cover the space with overlapping fields of view (with this relationship here shown by arrows). Because the aisles are symmetric, the data-driven perception models and graph topology trained on one primary aisle transfer directly to the structurally identical regions elsewhere. 4.3 Per… view at source ↗
Figure 5
Figure 5. Figure 5: Image-space speed for the nearest-neighbour baseline and our method: dense predictions across the image plane (left two panels; fixed orientation θ = 0°, white outline marks the training-data boundary) and per-trajectory error on held-out data (right two panels). Our method is smooth and perspective-consistent with low, uniform error, whereas the baseline is discontinuous where training data is sparse. ext… view at source ↗
Figure 6
Figure 6. Figure 6: Mean absolute error (MAE), relative to image dimensions, for bounding-box and speed predictions across all cameras and robots. method stays consistently low and outperforms both baselines. Errors peak at the most challenging cameras, such as 2 and 4 with their extreme perspectives or occlusions, and are lower for the smaller, more distinctive Jackal and Turtlebot. Errors generally remain below 0.5% of the … view at source ↗
Figure 7
Figure 7. Figure 7: View of three of the four robots at one instant of the end-to-end validation test, where each robot is shown from the perspective of the camera controlling it. Red lines indicate the path planned for each robot, and blue arrows show the detected in-image pose [PITH_FULL_IMAGE:figures/full_fig_p012_7.png] view at source ↗
read the original abstract

Off-board control of mobile robots from cameras embedded in the environment offers a practical path to scalable autonomy, moving sensing and compute off the robots. We extend this idea from the single-robot case to coordinated fleets in a real warehouse, driving multiple robots with only a distributed CCTV network and edge compute. The system operates entirely in image space over an uncalibrated, pixel-wise topological camera graph, enabling wide-area operation with flexible camera placement. A hierarchical planner selects a camera sequence per robot and plans its image-space motion through each view, coordinating robots with a prioritised-then-joint strategy and treating overlapping camera regions as shared resources held by one robot at a time to prevent collisions and deadlocks. We validate the approach in a real warehouse with four robots and 30 cameras across six 27 m aisles, reporting mission times and coordination statistics. To our knowledge, this is the first field demonstration of multi-robot planning and coordination using only an external camera network and off-board compute, with robots carrying no task-specific navigation hardware.

Editorial analysis

A structured set of objections, weighed in public.

Desk editor's note, referee report, simulated authors' rebuttal, and a circularity audit. Tearing a paper down is the easy half of reading it; the pith above is the substance, this is the friction.

Referee Report

0 major / 1 minor

Summary. The paper claims to be the first field demonstration of multi-robot planning and coordination using only an external camera network and off-board compute in a real warehouse. The system uses an uncalibrated pixel-wise topological camera graph for image-space planning and a prioritised-then-joint strategy to coordinate four robots with 30 cameras across six aisles, treating overlaps as shared resources to prevent collisions.

Significance. If the result holds, it has high practical significance for robotics applications in warehouses, as it allows using existing CCTV infrastructure for robot control without onboard navigation hardware. This could lower costs and increase flexibility. The field test provides valuable real-world data on mission times and coordination, supporting the feasibility of the approach. The absence of free parameters and the focus on implemented system are strengths.

minor comments (1)
  1. [Abstract] The abstract mentions reporting mission times and coordination statistics but does not present any specific numbers, failure rates, or baseline comparisons, which would help readers assess the results more readily.

Simulated Author's Rebuttal

0 responses · 0 unresolved

We thank the referee for the positive summary, recognition of the work's practical significance for warehouse robotics, and recommendation of minor revision. The referee's description of the system and contributions is accurate. No major comments were provided in the report.

Circularity Check

0 steps flagged

No significant circularity; systems paper with no derivation chain

full rationale

The paper describes an implemented multi-robot coordination system using an external uncalibrated camera network, with validation via field tests in a warehouse. No equations, fitted parameters, predictions, or first-principles derivations are present in the provided text. The central claim is an empirical demonstration rather than a mathematical result that could reduce to its inputs by construction. No load-bearing steps match any of the enumerated circularity patterns.

Axiom & Free-Parameter Ledger

0 free parameters · 1 axioms · 0 invented entities

The system rests on the domain assumption that a topological image-space graph suffices for collision-free multi-robot motion; no free parameters or new entities are introduced in the abstract.

axioms (1)
  • domain assumption A pixel-wise topological camera graph enables planning and coordination without metric calibration or 3D reconstruction.
    Stated directly in the abstract as the basis for wide-area operation with flexible camera placement.

pith-pipeline@v0.9.1-grok · 5717 in / 1184 out tokens · 33195 ms · 2026-06-28T00:41:18.784430+00:00 · methodology

discussion (0)

Sign in with ORCID, Apple, or X to comment. Anyone can read and Pith papers without signing in.

Reference graph

Works this paper leans on

28 extracted references

  1. [1]

    AAAI35(13), 11,220–11,227 (2021)

    Andreychuk, A., Yakovlev, K., Boyarski, E., Stern, R.: Improving continuous-time conflict based search. AAAI35(13), 11,220–11,227 (2021)

    2021

  2. [2]

    In: ICRA, vol

    Batalin, M.A., Sukhatme, G.S., Hattig, M.: Mobile robot navigation using a sensor network. In: ICRA, vol. 1, pp. 636–641. IEEE (2004)

    2004

  3. [3]

    In: IJCAI, pp

    Boyarski, E., Felner, A., Stern, R., Sharon, G., Tolpin, D., Betzalel, O., Shimony, E.: ICBS: improved conflict-based search algorithm for multi-agent pathfinding. In: IJCAI, pp. 740–746. AAAI Press, Buenos Aires, Argentina (2015)

    2015

  4. [4]

    IEEE Robot

    Buoso, D., Robinson, L., Averta, G., Torr, P., Franzmeyer, T., De Martini, D.: Select2plan: Training-free icl-based planning through vqa and memory retrieval. IEEE Robot. Autom. Lett.10(11), 11,267–11,274 (2025)

    2025

  5. [5]

    Chaumette, F., Hutchinson, S.: Visual servo control. i. basic approaches. IEEE Robot. Autom. Mag.13(4), 82–90 (2006)

    2006

  6. [6]

    Coulter, R.C.: Implementation of the pure pursuit path tracking algorithm. Tech. Rep. CMU-RI-TR-92-01, Carnegie Mellon University, Pittsburgh, PA (1992)

    1992

  7. [7]

    IEEE Trans

    Dixon, W., Dawson, D., Zergeroglu, E., Behal, A.: Adaptive tracking control of a wheeled mobile robot via an uncalibrated camera system. IEEE Trans. Syst. Man Cybern. B Cybern.31(3), 341–352 (2001)

    2001

  8. [8]

    In: IDAP, pp

    D¨ onmez, E., Kocamaz, A.F.: The eye-out-device multi-camera expansion for mobile robot control. In: IDAP, pp. 1–6. IEEE (2019)

    2019

  9. [9]

    IEEE Robot

    Kim, M., Kwon, Y., Lee, S., Yoon, S.e.: CCTV-informed human-aware robot nav- igation in crowded indoor environments. IEEE Robot. Autom. Lett.9(6), 5767– 5774 (2024) 16 Luke Robinson et al

    2024

  10. [10]

    IEEE Trans

    Liang, X., Wang, H., Liu, Y.H., Liu, Z., You, B., Jing, Z., Chen, W.: Purely image- based pose stabilization of nonholonomic mobile robots with a truly uncalibrated overhead camera. IEEE Trans. Robot.36(3), 724–742 (2020)

    2020

  11. [11]

    In: ICRA, vol

    Poduri, S., Sukhatme, G.S.: Constrained coverage for mobile sensor networks. In: ICRA, vol. 1, pp. 165–171. IEEE (2004)

    2004

  12. [12]

    In: AIR, pp

    Poornima, J., Krishnapuram, R., Bharatheesha, M., Amrutur, B., Sundaram, S.: Robust and scalable indoor robot localization based on fusion of infrastructure camera feeds and on-board sensors. In: AIR, pp. 1–7. Association for Computing Machinery, New York, NY, USA (2023)

    2023

  13. [13]

    In: CCC, pp

    Qingsong, L., Chaoli, W., Wenbin, N.: Tracking of nonholonomic control systems based on visual servoing feedback. In: CCC, pp. 459–463 (2007)

    2007

  14. [14]

    In: ECSA, vol

    Ravankar, A., Ravankar, A., Kobayashi, Y., Emaru, T.: Intelligent robot guidance in fixed external camera network for navigation in crowded and narrow passages. In: ECSA, vol. 1, p. 37. MDPI (2016)

    2016

  15. [15]

    Ravi, N., Gabeur, V., Hu, Y.T., Hu, R., Ryali, C., Ma, T., Khedr, H., R¨ adle, R., Rolland, C., Gustafson, L., Mintun, E., Pan, J., Alwala, K.V., Carion, N., Wu, C.Y., Girshick, R., Doll´ ar, P., Feichtenhofer, C.: Sam 2: Segment anything in images and videos (2024)

    2024

  16. [16]

    In: IROS (2023)

    Robinson, L., De Martini, D., Gadd, M., Newman, P.: Visual Servoing on Wheels: Robust Robot Orientation Estimation in Remote Viewpoint Control. In: IROS (2023)

    2023

  17. [17]

    In: ISER, pp

    Robinson, L., Gadd, M., Newman, P., Martini, D.D.: Robot-relay: Building-wide, calibration-less visual servoing with learned sensor handover networks. In: ISER, pp. 129–140. Springer (2023)

    2023

  18. [18]

    Robinson, L., Gadd, M., Newman, P., Martini, D.D.: Robot-relay: building-wide, calibration-less visual servoing with learned sensor handover networks. Auton. Robots50(1), 3 (2025)

    2025

  19. [19]

    Sharon, G., Stern, R., Felner, A., Sturtevant, N.R.: Conflict-based search for opti- mal multi-agent pathfinding. Artif. Intell.219, 40–66 (2015)

    2015

  20. [20]

    Sensors16(2), 195 (2016)

    Shim, J., Cho, Y.: A mobile robot localization via indoor fixed remote surveillance cameras. Sensors16(2), 195 (2016)

    2016

  21. [21]

    AIIDE1(1), 117–122 (2005)

    Silver, D.: Cooperative pathfinding. AIIDE1(1), 117–122 (2005)

    2005

  22. [22]

    Simoens, P., Dragone, M., Saffiotti, A.: The internet of robotic things: A re- view of concept, added value and applications. Int. J. Adv. Robot. Syst.15(1), 1729881418759,424 (2018)

    2018

  23. [23]

    Stern, R.: Multi-Agent Path Finding – An Overview, p. 96–115. Springer-Verlag, Berlin, Heidelberg (2022)

    2022

  24. [24]

    In: ADICS, pp

    Varghese, R., Sambath, M.: Yolov8: A novel object detection algorithm with en- hanced performance and robustness. In: ADICS, pp. 1–6 (2024)

    2024

  25. [25]

    In: IROS, pp

    Wagner, G., Choset, H.: M*: A complete multirobot path planning algorithm with performance bounds. In: IROS, pp. 3260–3267. IEEE, San Francisco, CA (2011)

    2011

  26. [26]

    IET Comput

    Whitaker, T.J., Cunningham, S.J., Bobda, C.: Decentralised indoor smart cam- era mapping and hierarchical navigation for autonomous ground vehicles. IET Comput. Vis.14(7), 462–470 (2020)

    2020

  27. [27]

    Yang, F., Su, H., Wang, C., Li, Z.: Adaptive and sliding mode tracking control for wheeled mobile robots with unknown visual parameters. Trans. Inst. Meas. Control40(1), 269–278 (2018)

    2018

  28. [28]

    Zhong, D., Robinson, L., Martini, D.D.: NeRFoot: Robot-footprint estimation for image-based visual servoing (2024)

    2024