Compressed Orientation Reachability Maps for Sub-Second Robotic Needle Entry Point Planning

Prabhu Nithin Gollapudi, Andreas Maier, Franziska Mathis-Ullrich, Water Zhao, Jia Wang, Siming BayerarXiv 2026 · arXiv

TL;DR

This paper introduces a compressed Orientation-Based Reachability Map (ORM) that slashes the time required for robotic needle entry point planning from nearly 30 seconds to under 0.4 seconds. By precomputing kinematic feasibility and storing it in a compact 83 MB bit-packed structure, the researchers enable real-time, intraoperative trajectory adjustments for robotic liver cancer treatments.

What Problem Does This Paper Address?

In percutaneous liver ablation, a robotic arm must insert a needle through the skin to destroy a tumor. A major bottleneck in this process is identifying a "feasible" entry point—one that the robot can actually reach given its joint limits and the surrounding obstacles.

Currently, planners rely on exhaustive Inverse Kinematics (IK) or randomized searches to verify every potential entry point. This process takes between 13 and 27 seconds per target. In a surgical setting where the liver moves due to respiration and tissue deforms under pressure, this delay is too long for responsive, real-time replanning. The challenge is to provide high-resolution kinematic validation without the massive computational and memory overhead usually associated with 6-Degree of Freedom (DoF) reachability maps.

Methods

The authors propose a "filter-and-verify" paradigm that shifts the heavy lifting to an offline phase:

  1. Offline ORM Construction: The robot's workspace is sampled at 188,873 discrete positions. For each position, the system exhaustively tests feasibility across a discretized orientation space (azimuth, elevation, and roll).
  2. Task-Specific Discretization: Instead of a generic grid, they use a parameterization tailored for needles: spherical coordinates for the approach vector plus a decoupled roll angle. This results in 186,624 orientations per point.
  3. Bit-Packed Compression: To avoid a "memory explosion," reachability flags are stored as individual bits in 64-bit integers. This compresses a massive 6D search space into a manageable 83 MB file.
  4. Runtime Pipeline: During surgery, the exhaustive IK search is replaced by a Just-In-Time (JIT) compiled pipeline. It performs $O(1)$ bitwise lookups to filter candidates, followed by linear interpolation for boundary smoothing and a final, selective IK check on the remaining "survivors."
Fig. 1. System architecture of the baseline and proposed ORM-based entry point planning pipelines. The proposed method shifts kinematic feasibility evaluation offline and performs online filtering through indexed reachability lookup.

Datasets

The study utilized a simulated model of a 6-DoF serial manipulator provided by Precision Medtech Co., Ltd.

  • Workspace: 188,873 discrete positions.
  • Evaluation: The system was tested against five representative target locations and a larger batch of 32 targets over 10 trials to measure amortized performance.
  • Note: Due to industry confidentiality, the specific robot model and joint limits were not disclosed, but the methodology is designed to be robot-agnostic.

Key Insights and Ideas

  • Task-Specific Geometry Matters: By recognizing that needle insertion is primarily about the approach vector, the authors could discretize the orientation space more efficiently than standard SE(3) voxel grids.
  • Memory Efficiency is the Enabler: A raw reachability map of this resolution would be gigabytes in size. Bit-packing allows the entire map to fit into a computer's cache or RAM (83 MB), making lookups near-instantaneous.
  • Reachability "Scores": The map doesn't just store "yes/no"; by counting feasible orientations at each point, it creates a reachability score that helps visualize the most "flexible" areas of the workspace.
Fig. 2. Reachability map visualized from three perspectives. Color intensity represents the reachability score (count of feasible orientations) mapped at each workspace coordinate location.

Key Results

  • Massive Speedup: Planning latency dropped from 13.5–26.9 seconds (baseline) to 0.289–0.349 seconds per target. This represents a 60–90× speedup.
  • High Fidelity: Despite the discretization, the ORM-filtered point clouds remained highly accurate compared to the ground-truth baseline, achieving a mean alpha-shape volume ratio of 0.976 and a mean convex hull overlap of 0.933.
  • Batch Efficiency: In large-scale evaluations, the system achieved a mean planning time of 0.188 seconds per target when processing batches, thanks to vectorized lookups.
Comparison showing the high overlap between the baseline (blue) and ORM-optimized (green) feasible entry points.

Limitations and Future Work

  • Discretization Artifacts: The authors noted small "omitted subsets" near the boundaries of the reachable workspace (see Figure 4 in the paper), which are inherent to grid-based discretization.
  • Static Environment: The current ORM assumes a static workspace. Future work involves integrating dynamic updates to handle intraoperative changes in the environment.
  • Deep Learning: The team plans to explore continuous reachability representations using Variational Autoencoders (VAEs) or Diffusion Models to eliminate boundary artifacts and further improve compression.
  • Clinical Validation: The next steps include moving from numerical simulations to hardware-in-the-loop validation using anatomical phantoms and eventually in-vivo studies.

Summary updated

Abstract

In the context of multi-stage trajectory planning for robotic percutaneous liver interventions, identifying a kinematically feasible entry point remains a critical computational bottleneck. Existing methods rely on exhaustive inverse kinematics or randomized searches, incurring delays of 13 to 27 seconds per target and limiting real-time intraoperative adaptation. This paper introduces an Orientation-Based Reachability Map (ORM) that precomputes the kinematic feasibility of a 6 Degrees of Freedom (DoF) serial manipulator across 188,873 workspace positions and encodes orientation feasibility into a compact, indexed 83 MB map. A Just-In-Time (JIT) compiled runtime pipeline replaces live optimization with near-instantaneous array lookups. Experimental evaluation demonstrates a reduction in planning latency from 13.5-26.9 s to 0.289-0.349 s per target while maintaining high geometric fidelity, achieving a mean convex hull volume overlap ratio of 0.933 and a mean alpha-shape volume ratio of 0.976 relative to the baseline reachable point cloud. The proposed ORM framework enables sub-second entry point planning and provides an efficient computational framework for robotic needle insertion in percutaneous liver ablation.
BibTeX
@misc{gollapudi2026compressed,
  author = {Prabhu Nithin Gollapudi and Andreas Maier and Franziska Mathis-Ullrich and Water Zhao and Jia Wang and Siming Bayer},
  title = {{Compressed Orientation Reachability Maps for Sub-Second Robotic Needle Entry Point Planning}},
  year = {2026},
  eprint = {2610.39879},
  archivePrefix = {arXiv},
  url = {https://arxiv.org/abs/2610.39879},
}
Figure