-00417A.svg){kind=icon}
Comprehensive developments for integrating physical AI into oncology clinical trials, by Claude Code, Cowork; with Assistance from ChatGPT and Google Gemini.
This repository provides production-ready configurations, validated pipelines, and integration guides for deploying robotic systems, digital twins, and embodied AI agents in oncology.
7/12: (Clinical Trial Funding Application v2.0) RFA-RM-27-001, Kawchak K. The application proposes a first-in-human, combined drug-device investigation of perioperative daraxonrasib and an eight-arm robotic pancreaticoduodenectomy."
7/7: (Clinical Trial Funding Application) RFA-RM-27-001. ChemicalQDevice respectfully submits this application to the NIH Director's Pioneer Award for "Daraxonrasib Phase 1 LLM-Directed Robotic Whipple in KRAS-Mutated PDAC."
7/1: v4.3.0 (Phase 1 PDAC IND: AI Generation, IND v1.0) v4.3.0 adds trial-ind/, a single-prompt Phase 1 PDAC IND that hastens the entire IND document-package process across the ReGARDD table of contents and ships 22 grayscale figures.
6/29: v4.2.0 (Phase 1 Pancreatic Cancer Trial Efficient LLM Document Generations, Paper v1.0) v4.2.0 adds trial-documents/, a single-prompt paper that hastens the entire Phase 1 process by generating every relevant documents.
6/23: v4.1.0 (Physical AI Pancreatic Whipple + Daraxonrasib Phase 2 Randomized Controlled Trial Protocol v1.1.0) v4.1.0 delivers the multicenter randomized Phase 2 follow-up: 220 participants randomized 1:1.
6/20: v4.0.0 (Physical AI Pancreatic Whipple + Daraxonrasib Phase 1 Protocol v1.0.0) v4.0.0 delivers the first substantial Physical AI clinical trial protocol: a Phase 1, first-in-human, combined IND/IDE study of on-premises LLM-directed robotic Whipple surgery with perioperative daraxonrasib.
5/10: v3.9.1 (Glioblastoma 1-Minute Variant Instructions) On-prem LLM controlled 1-minute glioblastoma resection simulation with 4 cooperating arms at mixed 1 kHz / 10 kHz force resolution and 16-iteration sweep - 12-file 1-minute set.
5/9: v3.9.0 (Glioblastoma Robotic Surgery Simulation Instructions) On-prem LLM controlled glioblastoma stereotactic resection simulation at 1 ms resolution for 1 hour with 64-iteration sweep - 17-file instruction set at competitions/instructions/
5/6: v3.8.0 (Patient Priority Final Paper) Patient Priority and Proposed U.S. Bills for Physical AI Oncology Clinical Trials - polished LaTeX manuscript at patients/paper/full-paper/final-paper populating the seven proposed federal bills.
5/3: v3.6.0 (Accelerated Patient Prediction Full Paper) Polished 70+ page LaTeX paper at new-trial/national-24-7-trial/paper/full-paper/ for "Accelerated Patient Prediction in Physical AI Oncology Clinical Trials: Four Comprehensive LLM Simulations"
5/1: v3.4.2 (National 24/7 Continuous RTCT) National 24/7 Continuous Real-Time Clinical Trial Simulation - FDA April 2026 Response - Indefinite-duration continuous trial spanning across 4 sites.
4/6: v3.4.0 (168-Hour Autonomous Sponsor Simulation) Fully Automated Sponsor: 7-Day Continuous Simulation with 168 Total Commits - 168 hourly Python scripts spanning across 7 total days.
4/4 PDF: v3.3.0 (Autonomous Sponsor Code Generation) Fully Automated Sponsor: Code Generation, Execution, and Paper Integration - Automated generation of 108 Python scripts (53 core agents, 24 hours.)
3/28 PDF: v3.0.0 (National Platform Paper) National Platform for Physical AI Oncology Trials - Comprehensive 186-page paper serving as an end-to-end resource for the pharmaceutical and industries.
3/24: v2.9.0 (Trial Site Documentation) Physical AI Oncology Clinical Trial Site Documentation - 11 LaTeX documents for California's first Physical AI oncology trial site: legislation drafts, regulations.
3/23: v2.8.0 (On-Demand Trial Simulation) 24-Hour On-Demand Physical AI Oncology Clinical Trial Simulation - Full 24-hour simulation of an autonomous, patient-centric oncology trial serving 168 patients.
3/20: v2.7.0 (Patient Journey Paper) A Cancer Patient's Journey Through a Regulated and Autonomous Physical AI Oncology Trial Illustration - Comprehensive paper documenting the patient journey
📄 3/20: v2.6.0 (Patient Journey) End-to-End Physical AI Oncology Clinical Trial Unification: Single-Patient Journey Orchestration - 10-stage patient journey for PAT-2026-0042 (58F, Stage IIIB NSCLC)
📄 3/18: v2.5.0 (Regulatory Adaptation) End-to-End Physical AI Oncology Clinical Trial Unification: Adaption of 21 CFR Part 312 - Investigational New Drug Application - Adaptation of 21 CFR Part 312 Regulation
📄 3/16: v2.4.0 (Regulatory Adaptation) End-to-End Physical AI Oncology Clinical Trial Unification: Adaption of 21 CFR Part 50 - Protection of Human Subjects - Adaptation of 21 CFR Part 50 Regulation
📄 3/12: v2.2.0 (Regulatory Guidance) End-to-End Physical AI Oncology Clinical Trial Unification. Comprehensive guidance adapted from prior ICH E6(R3), with Sections 1-4, Appendices A-C, and Glossary
📄 3/2: v2.1.0 (Patient Instructions) Patient Instructions: Physical AI Oncology Trials - Paper content documentation with page-by-page instructions, text diagrams, and quantitative patient data
📄 2/26: New Paper (USL) Unification Standard Level for Physical AI Oncology Trials. Standardizing and Evaluating Robot Unification Readiness for Multi-Site Clinical Trials. USL scores range from 1.0 to 10.0
v1.0.0 — First stable release. 51 Python modules (40,526 LOC), 69 documentation files, 28 examples, 5 CLI tools, and complete privacy/regulatory infrastructure. See V1_RELEASE.md for full release documentation.
This repository is complementary and open source, please implement code safely and responsibly. Intended audience: engineers building physical AI systems (robotics, ML, integration, and validation) for clinical trial settings.
# Clone the repository
git clone https://github.com/kevinkawchak/physical-ai-oncology-trials.git
cd physical-ai-oncology-trials
# Install base dependencies
pip install -r requirements.txt
# Verify framework availability
python scripts/verify_installation.py
# Detect available simulation frameworks
python unification/cross_platform_tools/framework_detector.pyThis release adds trial-ind/: a single-prompt Phase 1 Investigational
New Drug (IND) application, Phase 1 PDAC IND: AI Generation (IND v1.0), that hastens
the entire Phase 1 IND document-package process and ships 22 grayscale Mermaid figures,
each from a unique perspective, reproduced exactly in LaTeX. Authored by Kevin Kawchak (ChemicalQDevice), IND v1.0, DOI 10.5281/zenodo.21097442, July 1, 2026.
- Build pipeline (mermaid -> draft -> full -> final)
- Stage outputs
- IND at a glance
- trial-ind structure
%%{init: {'theme':'base','themeVariables':{'fontSize':'13px','lineColor':'#6C757D'}}}%%
flowchart LR
MP["Master prompt<br/>prompts/prompt-ind.md"]:::goal
S1["Stage 1 mermaid<br/>22 grayscale figures"]:::input
S2["Stage 2 draft-ind<br/>12 scaffolds + instructions"]:::input
S3["Stage 3 full-ind<br/>prose + TikZ + tables"]:::accent
S4["Stage 4 final-ind<br/>polished + zip"]:::goal
REL["Release v4.3.0<br/>README + releases + CHANGELOG"]:::proc
MP --> S1 --> S2 --> S3 --> S4 --> REL
classDef goal fill:#000000,stroke:#000000,stroke-width:1.5px,color:#FFFFFF
classDef proc fill:#3F3F3F,stroke:#000000,stroke-width:1.4px,color:#FFFFFF
classDef accent fill:#6C757D,stroke:#000000,stroke-width:1.3px,color:#FFFFFF
classDef input fill:#ECECEC,stroke:#3F3F3F,stroke-width:1.2px,color:#111111
| Stage | Directory | Output |
|---|---|---|
| Bootstrap | trial-ind/prompts, sub-prompts |
master prompt, output, 4 sub-prompts, READMEs |
| 1 mermaid | trial-ind/mermaid |
22 grayscale Mermaid figures (8-tone ramp) |
| 2 draft | trial-ind/draft-ind |
12 ReGARDD section scaffolds + zip |
| 3 full | trial-ind/full-ind |
12 full sections, 20 TikZ figures, 31 tables + zip |
| 4 final | trial-ind/final-ind |
polished sections, 22 figures, ~90 tables + zip (no publication dir) |
| Item | Value |
|---|---|
| Drug | Daraxonrasib (RMC-6236), oral RAS(ON) multi-selective inhibitor |
| Indication | Resectable / borderline KRAS G12-mutated PDAC; ECOG 0-1 |
| Design | Phase 1, first-in-human, combined IND / IDE; 3+3 dose finding |
| Dose levels | DL1 160 mg, DL2 220 mg, DL3 300 mg once daily; 28-day DLT |
| Device | On-premises LLM-directed eight-arm robotic Whipple (56 DOF, 640 channels) |
| Sample size | Up to 18 treated (about 36 screened), single academic site |
| Primary endpoints | 30-day device / procedure SAE rate; MTD / RP2D; task-completion feasibility |
| Document scale | 12 ReGARDD sections, 22 grayscale figures, ~90 full-width tables, ~10x the source paper |
trial-ind/
README.md build hub (badges, pipeline, milestones)
prompts/ prompt-ind.md (verbatim) + output-ind.md
sub-prompts/ prompt-1-mermaid .. prompt-4-final-ind
mermaid/ (Stage 1) 22 grayscale Mermaid figures + README + output
draft-ind/ (Stage 2) main.tex, indstyle.sty, references.bib, sections/, zip
full-ind/ (Stage 3) same set, 20 TikZ figures + 31 tables
final-ind/ (Stage 4) same set, polished, 22 figures (no publication subdirectory)
inputs/ ReGARDD IND template, FDA 1571 instructions, ReGARDD guidance, references.bib
This release adds trial-documents/: a single-prompt paper
(paper v1.0), Phase 1 Pancreatic Cancer Trial Efficient LLM Document Generations,
that shows how a repository based large language model, driven by one master prompt
that first writes and then executes a schedule of sub-prompts, hastens the entire
Phase 1 process by generating every relevant large trial document through a mermaid
to draft to full to final pipeline. Authored by Kevin Kawchak (ChemicalQDevice),
paper v1.0, DOI 10.5281/zenodo.21018646
June 29, 2026.
- Build pipeline (mermaid -> draft -> full -> final)
- Stage outputs
- The six acceleration targets
- trial-documents structure
%%{init: {'theme':'base','themeVariables':{'fontSize':'13px','lineColor':'#6C757D'}}}%%
flowchart LR
MP["Master prompt<br/>prompts/prompt-paper.md"]:::goal
S1["Stage 1 mermaid<br/>24 colored figures"]:::input
S2["Stage 2 draft-paper<br/>scaffold + instructions"]:::input
S3["Stage 3 full-paper<br/>prose + TikZ + tables"]:::accent
S4["Stage 4 final-paper<br/>polished + zip"]:::goal
REL["Release v4.2.0<br/>README + releases + CHANGELOG"]:::proc
MP --> S1 --> S2 --> S3 --> S4 --> REL
classDef goal fill:#8B2E3F,stroke:#000000,stroke-width:1.5px,color:#FFFFFF
classDef proc fill:#2F5D7C,stroke:#000000,stroke-width:1.4px,color:#FFFFFF
classDef accent fill:#D08770,stroke:#000000,stroke-width:1.3px,color:#111111
classDef input fill:#BFD7EA,stroke:#2F5D7C,stroke-width:1.2px,color:#111111
| Stage | Directory | Output |
|---|---|---|
| Bootstrap | trial-documents/prompts, sub-prompts |
master prompt, output, 4 sub-prompts, READMEs |
| 1 mermaid | trial-documents/mermaid |
24 colored Mermaid figures (5-step palette) |
| 2 draft | trial-documents/draft-paper |
8 section scaffolds + zip |
| 3 full | trial-documents/full-paper |
8 full sections, 24 TikZ figures, 6 tables + zip |
| 4 final | trial-documents/final-paper |
polished sections + zip (no publication dir) |
| # | Document target | Gate | Binding clock |
|---|---|---|---|
| 1 | Initial IND and IRB package | Hard | 30-day FDA review; IRB calendar |
| 2 | Protocol amendments + synchronized consent | Hard | IRB approval; serial-revision loss |
| 3 | Cohort-review packages after safety data mature | Protocol-defined | DLT observation window |
| 4 | Complete clinical-hold response | Hard | 30-day FDA review of a complete response |
| 5 | Phase 2-to-3 briefing package and Phase 3 protocol | Decision | EOP2 scheduling |
| 6 | Pivotal CSR and NDA/BLA modules after database lock | Decision/filing | Database lock; RTOR staging |
trial-documents/
README.md build hub (badges, pipeline, milestones)
prompts/ prompt-paper.md (verbatim) + output-paper.md
sub-prompts/ prompt-1-mermaid .. prompt-4-final-paper
mermaid/ (Stage 1) 24 colored Mermaid figures + README + output
draft-paper/ (Stage 2) main.tex, paperstyle.sty, references.bib, sections/, zip
full-paper/ (Stage 3) same set, 24 TikZ figures + 6 tables
final-paper/ (Stage 4) same set, polished (no publication subdirectory)
inputs/ llm-adoption template + references.bib
research/ document-types (2) + industry-workflow (2) AI sources
This release adds trial-phase-2/: the multicenter randomized
Phase 2 follow-up to the Phase 1 protocol. With the Phase 1 protocol having
established the daraxonrasib recommended Phase 2 dose (RP2D, 300 mg once daily) and
the feasibility and safety of the on-premises LLM-directed eight-arm robotic
Whipple, genuine clinical equipoise now exists, so this study randomizes 220
participants 1:1 across eight high-volume academic centers. Authored by Kevin Kawchak (ChemicalQDevice), protocol
v1.1.0, DOI 10.5281/zenodo.20807027,
June 23, 2026.
- Build pipeline (mermaid -> draft -> full -> final -> publication)
- Stage outputs
- Protocol at a glance (Phase 2)
- trial-phase-2 structure
%%{init: {'theme':'base','themeVariables':{'fontSize':'13px','lineColor':'#6B6B6B'}}}%%
flowchart LR
MP["Master prompt<br/>prompts/prompt-protocol.md"]:::goal
S1["Stage 1 mermaid<br/>24 colored figures"]:::light
S2["Stage 2 draft<br/>bracketed scaffold"]:::light
S3["Stage 3 full<br/>22 TikZ figures, 11 tables"]:::mid
S4["Stage 4 final<br/>polished + zip"]:::mid
PUB["Publication<br/>author edits, paper URL dir"]:::goal
REL["Release v4.1.0<br/>CHANGELOG + releases + README"]:::goal
MP --> S1 --> S2 --> S3 --> S4 --> PUB --> REL
classDef light fill:#F5F5F5,stroke:#800020,stroke-width:1.2px,color:#111111
classDef mid fill:#6B6B6B,stroke:#111111,stroke-width:1.4px,color:#FFFFFF
classDef goal fill:#800020,stroke:#000000,stroke-width:1.6px,color:#FFFFFF
| Stage | Directory | Output |
|---|---|---|
| Bootstrap | trial-phase-2/prompts, sub-prompts |
master prompt, 4 sub-prompts, READMEs |
| 1 mermaid | trial-phase-2/mermaid |
24 colored Mermaid figures (Burgundy #800020 palette) |
| 2 draft | trial-phase-2/draft-protocol |
13 NIH section scaffolds + zip |
| 3 full | trial-phase-2/full-protocol |
13 full sections, 22 TikZ, 11 tables + zip |
| 4 final | trial-phase-2/final-protocol |
polished sections + zip |
| publication | trial-phase-2/final-protocol/publication |
author-edited paper URL directory + zip |
| Element | Value |
|---|---|
| Design | Phase 2, multicenter (8 centers), randomized 1:1, parallel-group, controlled, open-label with BICR |
| Arm A | Daraxonrasib RP2D (300 mg once daily) + on-premises LLM-directed eight-arm robotic Whipple |
| Arm B | Modified FOLFIRINOX + institutional-standard high-volume pancreaticoduodenectomy |
| Primary endpoint | Progression-free survival; HR 0.60; 85% power; two-sided alpha 0.05; about 140 events; one group-sequential interim |
| Key secondary (hierarchical) | OS; R0 rate; ISGPS grade B/C fistula; major pathologic response; ctDNA clearance |
| Device readiness | Phase 0 USL >= 8.0; >= 5000 sims; >= 3 frameworks; sim-to-real < 1.5 mm / < 0.4 N; fleet harmonization |
| Funding | Patient-Aligned Co-Investment Facility behind a capital firewall (21 CFR part 54; H.R. 9510 VVUQ) |
trial-phase-2/
├── README.md # build hub (v1.1.0)
├── prompts/ # prompt-protocol.md (master) + output-protocol.md
├── sub-prompts/ # prompt-1-mermaid .. prompt-4-final-protocol
├── mermaid/ # Stage 1: 24 colored Mermaid figures (#800020 palette)
├── draft-protocol/ # Stage 2: 13 NIH section scaffolds + zip
├── full-protocol/ # Stage 3: full sections, 22 TikZ, 11 tables + zip
├── final-protocol/ # Stage 4: polished sections + zip
│ └── publication/ # author-edited paper URL directory (the paper)
├── template/ # paper template (recolored #800020)
├── nih-protocol/ # NIH-FDA Phase 2/3 IND/IDE template grounding
├── inputs/ # main documents + Phase 1 predicate (grounding)
└── research/ # Phase 2 evidence base and background
This release adds trial-protocol/: the first substantial
Physical AI clinical trial protocol in this repository. It is a Phase 1,
first-in-human, combined IND/IDE study of on-premises LLM-directed robotic
pancreaticoduodenectomy (the Whipple procedure) with perioperative
daraxonrasib (RMC-6236) in KRAS-mutated pancreatic ductal adenocarcinoma. Authored by Kevin Kawchak (ChemicalQDevice), DOI 10.5281/zenodo.20780121,
June 20, 2026.
- Build pipeline (mermaid -> draft -> full -> final)
- Stage outputs and commit counts
- Protocol at a glance
- trial-protocol structure
%%{init: {'theme':'base','themeVariables':{'fontSize':'13px','lineColor':'#6C757D'}}}%%
flowchart LR
MP["Master prompt<br/>prompts/prompt-protocol.md"]:::goal
S1["Stage 1 mermaid<br/>25 colored figures"]:::light
S2["Stage 2 draft<br/>bracketed scaffold"]:::light
S3["Stage 3 full<br/>20 TikZ figures, 11 tables"]:::mid
S4["Stage 4 final<br/>polished + zip"]:::goal
REL["Release v4.0.0<br/>CHANGELOG + releases + README"]:::goal
MP --> S1 --> S2 --> S3 --> S4 --> REL
classDef light fill:#FFFFFF,stroke:#111111,stroke-width:1px,color:#111111
classDef mid fill:#6C757D,stroke:#111111,stroke-width:1.2px,color:#FFFFFF
classDef goal fill:#00417A,stroke:#000000,stroke-width:1.5px,color:#FFFFFF
| Stage | Directory | Output | Commits |
|---|---|---|---|
| Bootstrap | trial-protocol/prompts, sub-prompts |
master prompt, 4 sub-prompts, READMEs | per file |
| 1 mermaid | trial-protocol/mermaid |
25 colored Mermaid figures | 28 |
| 2 draft | trial-protocol/draft-protocol |
13 NIH section scaffolds + zip | 21 |
| 3 full | trial-protocol/full-protocol |
13 full sections, 20 TikZ, 11 tables + zip | 21 |
| 4 final | trial-protocol/final-protocol |
polished sections + zip | 21 |
| Element | Value |
|---|---|
| Design | Phase 1, first-in-human, open-label, single-arm, combined IND/IDE |
| Drug arm | Daraxonrasib (RMC-6236), RAS(ON) multi-selective inhibitor, 3+3 (160 / 220 / 300 mg) |
| Device arm | Eight-arm LLM-directed robotic Whipple; 56 DOF; 640 sensor channels; 3 N / 18 N force caps; 3 ms E-stop |
| Population | KRAS G12 PDAC, ECOG 0-1, resectable / borderline-resectable; up to n = 18 |
| Primary endpoints | 30-day device/procedure serious-AE incidence; MTD/RP2D; unsafe-conversion-free task completion |
| Safety | 5-vessel no-fly gate; VVUQ ten-gate; hash-chained audit trail (21 CFR part 11); USL |
trial-protocol/
├── README.md # build hub (v4.0.0)
├── prompts/ # prompt-protocol.md (master) + output-protocol.md
├── sub-prompts/ # prompt-1-mermaid .. prompt-4-final-protocol
├── mermaid/ # Stage 1: 25 colored Mermaid figures
├── draft-protocol/ # Stage 2: 13 NIH section scaffolds + zip
├── full-protocol/ # Stage 3: full sections, 20 TikZ, 11 tables + zip
├── final-protocol/ # Stage 4: polished sections + zip
├── template/ # paper template (recolored #00417A)
├── nih-protocol/ # NIH-FDA IND/IDE template (10 chunks)
├── inputs/ # 3 main documents + author_works.bib
└── research/ # 2026 Physical AI FDA + oncology-strategy markdowns
4-Arm Sensor Streams Per-Arm XYZ Commands 1-Min vs 1-Hr Compare
(50 ch/arm x 4 arms, ---> (per-arm phase- --> (on-prem LLM,
200 ch total at mixed conditioned 1 kHz 4-round tournament)
1 kHz + 10 kHz force) with 5 ms e-stop)
+-----------------------+ +------------------------+ +----------------+
| Arm 1 hyb u-w-p cut | -> | Per-arm x, y, z, q, |-> | Quality 0.40 |
| Arm 2 bipolar coag | | linear_vel up to | | Time 0.25 |
| Arm 3 suction collect | | 1,000 mm/s, force | | Cost 0.20 |
| Arm 4 iMRI + 5-ALA | | clamp 5 N/arm, tool, | | Safety 0.10 |
| 1 kHz heartbeat bus | | 7-state command enum | | PtExp 0.05 |
| 12 N cumulative cap | | + heartbeat watchdog | | TrueSkill mu/s |
+-----------------------+ +------------------------+ +----------------+
| | |
v v v
+-----------------------+ +-----------------------+ +----------------+
| NeuroSpeed 1.0 (2030) | | 4-phase 60s timeline | | Compare vs |
| 4 arms x 7 DOF, 28 | | P1 dural 0-5s, P2 | | v3.9.0 1-hr |
| DOF total, 0.1 mm RMS | | bulk 5-45s @ 800 | | ROSA ONE Brain |
| at 1,000 mm/s, 5 ms | | mm cubed per s, P3 | | v3.0 baseline |
| e-stop, 800 mm cubed | | margin 45-55s, P4 | | + manual human |
| per s peak via hybrid | | hemostasis 55-60s | | (no published |
| u-w-p removal | | (pre-op precomputed) | | 1-min human) |
+-----------------------+ +-----------------------+ +----------------+
| | |
v v v
+-----------------------------------------------------------------------------+
| v3.9.1: 12-file 1-minute variant instruction set at |
| competitions/instructions/one_minute_variant/ for a future Claude Code Opus |
| 4.7 1M Max session to author the simulation across 7 sequential commits in |
| 1 PR. L1 (20 Hz) + L2 (1 Hz) + L3 (per-phase) + events committed at 510 KB |
| per iteration; 16 iterations total 8.2 MB committed within 10 MB cap. L0 |
| raw of 26 MB per iteration archived to Zenodo (416 MB total, free 50 GB |
| tier). Inherits parent v3.9.0 instructions verbatim except for 4-arm and |
| 1-minute-specific overrides; nothing in glioblastoma-1hr-trial/ is touched. |
+-----------------------------------------------------------------------------+
Sensor Stream XYZ Commands Comparison Agent
(50 channels @ 1 kHz) ---> (phase-conditioned ---> (on-prem LLM,
1 kHz / 100 Hz) skill rating)
+-----------------------+ +-----------------------+ +----------------+
| Joints x 18 channels | -> | x, y, z, qw, qx, qy, |-> | Quality 0.40 |
| EE pose x 7 channels | | qz, vel, force_clamp, | | Time 0.25 |
| EE force x 6 channels | | tool, command_state | | Cost 0.20 |
| Nav dev x 3 channels | | + 2 metadata | | Safety 0.10 |
| Tool flags x 6 chan. | | ~2.73M commands/hr | | PtExp 0.05 |
| Safety enums x 2 ch. | | Parquet 90 MB | | TrueSkill mu/s |
| + 8 metadata fields | +-----------------------+ +----------------+
| 3.6M ticks/hr | | |
| Parquet 60 MB | v v
+-----------------------+ +-----------------------+ +----------------+
| | ROSA ONE Brain v3.0 | | Tournament 8 |
v | firmware 3.1.4 | | rounds: this |
+-----------------------+ | 6 DOF, 0.5 mm RMS, | | project vs. |
| Layer 1: generators | | 50 mm/s max linear | | prior version |
| Layer 2: 7 commits in | | velocity, IEC 80601- | | snapshots, |
| one PR with per- | | 2-77 force limits, | | NeuroMate, |
| commit budgets | | 21 CFR 50.30 task- | | Brainlab Cirq, |
| Layer 3: per-file | | order lifecycle | | Modus V, |
| chunking caps | +-----------------------+ | Mazor X, human |
+-----------------------+ +----------------+
| | |
v v v
+-----------------------------------------------------------------------------+
| v3.9.0: 17-file instruction set at competitions/instructions/ for a future |
| Claude Code Opus 4.7 1M Max session to author the simulation across 7 |
| sequential commits in 1 PR. ASCII and Mermaid diagrams replace SVG for |
| high-frequency series. SHA-256 manifested release snapshot enables future |
| release-vs-release comparisons via the on-prem LLM comparison agent. |
+-----------------------------------------------------------------------------+
Four LLM Simulations Paper Template Future Pass
(Inputs) ---> (this PR) ---> (Population)
+-------------------+ +---------------------+ +------------+
| Sim 1: 24/7 RTCT | ---> | main.tex | ---> | 70+ pages |
| (84 hours x 7 | ---> | new_paper.sty | ---> | Color |
| files, 4 sites, | | references.bib (35) | | diagrams |
| 116 robots) | | abstract.tex | | added in a |
| Sim 2: Patient | ---> | introduction.tex | ---> | second |
| Journey (10 | | methods.tex (prose) | | author- |
| stages, 1 patient)| | results.tex (4 sims)| | driven |
| Sim 3: 24h | ---> | discussion.tex | ---> | pass |
| Sponsor (24 .py, | | limitations.tex | | |
| 53 agents, 75 dx) | | conclusions.tex | | Senior |
| Sim 4: 168h, 7d | ---> | back_matter.tex | ---> | author |
| (525 dx, 168 .py, | | LaTeX_Source_Files | | white- |
| 7 PRs, Core i5) | | .zip (Overleaf) | | space pass |
+-------------------+ +---------------------+ +------------+
| | |
v v v
+-----------------------------------------------------------------------------+
| v3.5.0: bracketed instructions naming exact directories and file paths, |
| ASCII diagrams to embed verbatim, individual patient and robot examples, |
| and DOIs and clickable URLs for every reference (GitHub + Zenodo). |
+-----------------------------------------------------------------------------+
physical-ai-oncology-trials/
├── README.md
├── V1_RELEASE.md
├── LICENSE
├── requirements.txt
│
├── trial-ind/ # ★ Phase 1 PDAC IND: AI Generation (v4.3.0)
│ ├── README.md # build hub (IND v1.0 / repo v4.3.0)
│ ├── prompts/ # prompt-ind.md (master, verbatim) + output-ind.md
│ ├── sub-prompts/ # prompt-1-mermaid .. prompt-4-final-ind
│ ├── mermaid/ # Stage 1: 22 grayscale Mermaid figures + README + output
│ ├── draft-ind/ # Stage 2: 12 ReGARDD section scaffolds + zip
│ ├── full-ind/ # Stage 3: 12 full sections, 20 TikZ, 31 tables + zip
│ ├── final-ind/ # Stage 4: polished, 22 figures, ~90 tables + zip (no publication dir)
│ └── inputs/ # ReGARDD IND template, FDA 1571 instructions, references.bib
│
├── trial-documents/ # ★ Phase 1 PDAC Efficient LLM Document Generations Paper (v4.2.0)
│ ├── README.md # build hub (paper v1.0 / repo v4.2.0)
│ ├── prompts/ # prompt-paper.md (verbatim) + output-paper.md
│ ├── sub-prompts/ # 4 generated stage sub-prompts
│ ├── mermaid/ # Stage 1: 24 colored Mermaid figures (5-step palette)
│ ├── draft-paper/ # Stage 2: 8 section scaffolds + zip
│ ├── full-paper/ # Stage 3: 8 full sections, 24 TikZ, 6 tables + zip
│ ├── final-paper/ # Stage 4: polished sections + zip
│ │ └── publication/ # author-edited paper URL directory (the paper)
│ ├── inputs/ # llm-adoption template + references.bib
│ └── research/ # document-types + industry-workflow AI sources
│
├── trial-phase-2/ # ★ Physical AI Whipple + Daraxonrasib Phase 2 Randomized Controlled Trial Protocol (v4.1.0)
│ ├── README.md # build hub (v1.1.0 / repo v4.1.0)
│ ├── prompts/ # Phase 2 master prompt (verbatim) + output
│ ├── sub-prompts/ # 4 generated stage sub-prompts
│ ├── mermaid/ # Stage 1: 24 figures recolored to #800020
│ ├── draft-protocol/ # Stage 2: 13 NIH section scaffolds + zip
│ ├── full-protocol/ # Stage 3: full sections, 22 TikZ, 11 tables + zip
│ ├── final-protocol/ # Stage 4: polished sections + zip
│ │ └── publication/ # author-edited paper URL directory (the paper)
│ ├── template/ # paper template (recolored #800020)
│ ├── nih-protocol/ # NIH-FDA Phase 2/3 IND/IDE template grounding
│ ├── inputs/ # main documents + Phase 1 predicate (grounding)
│ └── research/ # Phase 2 evidence base and background
│
├── trial-protocol/ # ★ Physical AI Whipple + Daraxonrasib Phase 1 Protocol (v4.0.0)
│ ├── README.md # build hub
│ ├── prompts/ # master prompt (verbatim) + full output
│ ├── sub-prompts/ # 4 generated stage sub-prompts
│ ├── mermaid/ # Stage 1: 25 colored Mermaid figures
│ ├── draft-protocol/ # Stage 2: 13 NIH section scaffolds + zip
│ ├── full-protocol/ # Stage 3: full sections, 20 TikZ, 11 tables + zip
│ ├── final-protocol/ # Stage 4: polished sections + zip
│ ├── template/ # paper template (recolored #00417A)
│ ├── nih-protocol/ # NIH-FDA IND/IDE template (10 chunks)
│ ├── inputs/ # 3 main documents + author_works.bib
│ └── research/ # 2026 Physical AI FDA + oncology-strategy markdowns
│
├── sponsor/ # ★ Fully Automated Sponsor (v3.3.0)
│ ├── input_files/ # Sponsor playbook and organization inputs (16 files)
│ │ ├── README.md # Cross-document alignment and processing notes
│ │ ├── sponsor_01-08_*.md # End-to-End Sponsor Playbook (8 chunks)
│ │ └── org_01-07_*.md # Sponsor Organization (7 chunks)
│ ├── paper/ # Complete Autonomous Sponsor Paper (v3.2.0)
│ │ ├── main.tex # Main document (18 sections + 6 appendices)
│ │ ├── sponsor_paper.sty # Style file (adapted from arxiv.sty, CC BY 4.0)
│ │ ├── references.bib # Bibliography (48+ entries with DOIs/URLs)
│ │ ├── README.md # Paper documentation and compilation guide
│ │ └── sections/ # 19 section .tex files (complete paper content)
│ ├── final_paper/ # ★ Final Paper with Code Generations (v3.3.0)
│ │ ├── main.tex # Updated document with execution results
│ │ ├── sections/ # 19 updated .tex files
│ │ ├── README.md # Comprehensive documentation
│ │ ├── scripts/ # 108 generated Python scripts (v3.3.0)
│ │ │ ├── run_sponsor_simulation.py # Master 24-hour simulation runner
│ │ │ ├── generate_all_diagrams.py # 75 text diagram generator
│ │ │ ├── sponsor_server/ # FastAPI sponsor control server (15 files)
│ │ │ ├── hourly/ # 24 hourly sponsor generators + JSON output
│ │ │ ├── diagrams/ # 75 ASCII text diagrams (3 perspectives)
│ │ │ ├── coordination/ # Agent event bus, escalation, gates
│ │ │ ├── safety/ # Robotic safety workflows
│ │ │ ├── dashboard/ # Terminal dashboard and reports
│ │ │ ├── core_agents/ # 53 core agent implementations
│ │ │ └── output/ # Simulation results and reports
│ │ └── 168_hours/ # ★ 168-Hour Simulation (v3.4.0)
│ │ ├── README.md # Simulation overview and statistics
│ │ ├── run_168h_simulation.py # Master 168-hour simulation runner
│ │ ├── day_01/ - day_07/ # 7 day directories (24 hours each)
│ │ │ ├── hourly/ # 24 sponsor_hour_XXX.py + JSON output
│ │ │ ├── diagrams/ # 75 text diagrams per day (72 hourly + 3 cumulative)
│ │ │ └── output/ # Day summary JSON
│ │ └── instructions/ # Real-time execution instructions
│ │ ├── rtx_4090_openclaw/ # RTX 4090 setup (Linux, macOS, Windows)
│ │ ├── mac_mini_m4_pro_openclaw/ # M4 Pro setup (Linux, macOS, Windows)
│ │ └── core_i5_6200u_4gb/ # Core i5-6200U 4GB setup (Windows 10 Pro)
│ └── template/ # Autonomous Sponsor Paper Template (v3.1.0)
│ ├── main.tex # Template document (18 sections + appendices)
│ ├── sponsor_paper.sty # Style file (adapted from arxiv.sty, CC BY 4.0)
│ ├── references.bib # Bibliography (48 entries with DOIs/URLs)
│ ├── orcid_icon.png # ORCID icon for author attribution
│ ├── README.md # Template documentation and processing guide
│ └── sections/ # 19 section .tex files with processing instructions
│
├── new-trial/ # ★ 24-Hour On-Demand Trial Simulation (v2.8.0)
│ ├── README.md # Simulation overview and results
│ ├── psl_framework.md # PSL scoring framework definition
│ ├── site_specification.md # Facility and staffing specifications
│ ├── format_comparison.md # On-demand vs. traditional comparison
│ ├── prompts.md # v2.8.0 development prompt
│ ├── hour-00/ through hour-23/ # 24 hourly simulation directories
│ │ ├── hour_XX_simulation.md # Master simulation log
│ │ ├── hour_XX_robot_logs.md # Per-robot telemetry
│ │ ├── hour_XX_patient_records.md # Patient vitals and records
│ │ ├── hour_XX_psl_scores.md # PSL scores for all 10 robots
│ │ ├── hour_XX_diagram_facility.txt # Facility layout diagram
│ │ ├── hour_XX_diagram_patient_flow.txt # Patient flow diagram
│ │ └── hour_XX_diagram_robot_status.txt # Robot status timeline
│ ├── final-commit/ # Error review and 24-hour summaries
│ │ ├── final_error_review.md # Consistency check
│ │ ├── final_24h_summary.md # Performance summary
│ │ ├── final_psl_cumulative.md # PSL trajectory analysis
│ │ └── final_diagram_*.txt # Summary diagrams (3 files)
│ ├── national-24-7-trial/ # ★ National 24/7 Continuous RTCT (v3.4.2)
│ │ ├── README.md # Continuous trial overview and FDA mapping
│ │ ├── FDA-April-2026/ # Source FDA news release (28 Apr 2026)
│ │ ├── Background-A/ # Deep research chunk set A (17 bib entries)
│ │ ├── Background-B/ # Deep research chunk set B (9 bib entries)
│ │ ├── hour-00/ through hour-55/ # Hourly folders, 7 files each
│ │ ├── extra-hours/ # hour-56 through hour-83 (approximated)
│ │ └── paper/ # Paper Template (v3.5.0)
│ │ ├── main.tex # Document skeleton + global formatting brief
│ │ ├── new_paper.sty # Style file (arxiv-derived, CC BY 4.0)
│ │ ├── references.bib # 35 entries with DOIs and clickable URLs
│ │ ├── orcid_icon.png # Title-page ORCID hyperlink asset
│ │ ├── README.md # Paper-template documentation
│ │ ├── LaTeX_Source_Files.zip # Overleaf-ready archive
│ │ ├── sections/ # 8 section .tex files
│ │ └── full-paper/ # ★ Polished Full Paper (v3.6.0)
│ │ ├── main.tex # Polished document with stronger formatting
│ │ ├── new_paper.sty # Updated style with displaywidowpenalty
│ │ ├── references.bib # 35 entries + 2 added Zenodo refs
│ │ ├── orcid_icon.png # Title-page ORCID hyperlink asset
│ │ ├── README.md # DOI badges + ASCII diagrams + tables
│ │ ├── LaTeX_Source_Files.zip # Overleaf-ready archive
│ │ └── sections/ # 8 section .tex files (final prose)
│ └── site/ # Trial Site Documentation (v2.9.0)
│ ├── README.md # Site documentation overview
│ ├── 01-legislation-authorization/ # SB 1042 authorization act
│ ├── 02-legislation-patient-rights/ # AB 2847 patient rights act
│ ├── 03-legislation-data-transparency/ # SB 892 data protection act
│ ├── 04-city-regulations/ # SF municipal code update
│ ├── 05-state-regulations/ # CA Title 22 Chapter 14
│ ├── 06-national-regulations/ # FDA compliance guide
│ ├── 07-building-code/ # Facility construction standards
│ ├── 08-premises-code/ # Site safety and access
│ ├── 09-parking-transportation/ # Parking and transit standards
│ ├── 10-site-operations/ # Activation and SOPs
│ ├── 11-emergency-preparedness/ # Emergency response plan
│ ├── all-documents/ # Combined 11-document source
│ │ ├── all_documents.tex # Full combined LaTeX source
│ │ └── all_documents_chunk/ # Chunked into 11 files (v2.9.1)
│ │ └── README.md # Reconstruction instructions
│ └── zips/ # LaTeX source archives (12 zips)
│
├── patient-journey/ # ★ Single-Patient Journey Orchestration (v2.6.0)
│ ├── patient_state.py # Central data model (10 enums, 14 dataclasses)
│ ├── stage_01_prescreening.py # Stage 1: Pre-Screening & Referral Intake
│ ├── stage_02_enrollment.py # Stage 2: Enrollment & Informed Consent
│ ├── stage_03_digital_twin.py # Stage 3: Digital Twin Construction
│ ├── stage_04_robot_qualification.py# Stage 4: Robot Qualification
│ ├── stage_05_surgery.py # Stage 5: Robot-Assisted Surgery
│ ├── stage_06_recovery.py # Stage 6: Post-Operative Recovery
│ ├── stage_07_immunotherapy.py # Stage 7: Immunotherapy Treatment
│ ├── stage_08_federation.py # Stage 8: Federated Learning
│ ├── stage_09_surveillance.py # Stage 9: Long-Term Surveillance
│ ├── stage_10_closeout.py # Stage 10: Trial Closeout
│ ├── master_journey.py # Master orchestrator (all 10 stages)
│ ├── diagrams/ # 30 ASCII progress diagrams (3 x 10)
│ ├── deliverables/ # Charts, tables, FDA analysis, guidance
│ ├── paper/ # ★ Patient Journey Paper (v2.7.0)
│ │ ├── patient_journey_paper.tex # LaTeX source
│ │ ├── patient_journey_paper_chunk/ # Chunked into 3 files (v2.9.1)
│ │ │ └── README.md # Reconstruction instructions
│ │ ├── patient_journey_paper.pdf # Compiled PDF (compile from .tex)
│ │ ├── Latex_Source_Code.zip # Source archive
│ │ ├── arxiv.sty # Style file
│ │ ├── orcid_icon.png # ORCID icon
│ │ ├── README.md # Paper documentation
│ │ └── template/ # Formatting template
│ └── prompts.md # Development prompts archive
│
├── patients/ # ★ Patient Instructions (v2.1.0)
│ ├── patient_robot_instructions_fixed.tex # LaTeX source (10 robot categories)
│ ├── patient_robot_instructions_fixed_chunk/ # Chunked into 2 files (v2.9.1)
│ │ └── README.md # Reconstruction instructions
│ ├── README.md # Paper content, instructions, text diagrams
│ ├── paper/ # ★ Patient Priority Paper Template (v3.7.0)
│ │ ├── README.md # Template documentation, AVAILABLE DIRECTORIES
│ │ ├── main.tex # Document skeleton + global formatting brief
│ │ ├── patient_priority.sty # Style file (adapted from prior templates)
│ │ ├── patient_priority.bib # 56 entries with DOIs + clickable URLs (biber)
│ │ ├── LaTeX_Source_Files.zip # Overleaf-ready archive
│ │ ├── Deep-Research-1/ # AI and patient control evidence (3 chunks)
│ │ ├── Deep-Research-2/ # Layered legal-stack baseline (4 chunks)
│ │ ├── Deep-Research-3/ # Oncology trial laws + patient control (4 parts)
│ │ ├── sections/ # 15 template section .tex files (v3.7.0)
│ │ │ ├── abstract.tex
│ │ │ ├── introduction.tex
│ │ │ ├── patient_priority.tex
│ │ │ ├── hr_4501_patient_self_selection.tex
│ │ │ ├── hr_4502_robot_humanoid_choice.tex
│ │ │ ├── hr_4503_procedural_modification.tex
│ │ │ ├── hr_4504_error_reduction.tex
│ │ │ ├── hr_4505_realtime_sponsor.tex
│ │ │ ├── hr_4506_american_leadership.tex
│ │ │ ├── hr_4507_data_self_custody.tex
│ │ │ ├── implementation_metrics.tex
│ │ │ ├── discussion.tex
│ │ │ ├── limitations_future.tex
│ │ │ ├── conclusions.tex
│ │ │ └── back_matter.tex
│ │ └── full-paper/ # ★ Patient Priority Full Paper (v3.8.0)
│ │ ├── README.md # Full paper documentation + DOI badges + ASCII diagram
│ │ ├── main.tex # 7-Bill consolidated document entry point
│ │ ├── patient_priority.sty # Polished style with widow/orphan + flushbottom
│ │ ├── patient_priority.bib # 47 entries, single canonical URL each (biber)
│ │ ├── LaTeX_Source_Files.zip # Overleaf-ready archive
│ │ └── sections/ # 8 final-prose section .tex files
│ │ ├── hr_9501_patient_self_selection.tex # Adaption of 21 CFR 50, FDA DCT, 42 USC 300gg-8
│ │ ├── hr_9502_robot_humanoid_choice.tex # Revision of CA AB 2847, FDA AI Draft 2025
│ │ ├── hr_9503_procedural_modification.tex # Adaption of OHRP Broad Consent, Cures Act
│ │ ├── hr_9504_error_reduction.tex # Adaption of HTI-1 DSI, FDA AI Draft 2025
│ │ ├── hr_9505_realtime_sponsor.tex # Revision of FDA RTCT, FDA DCT
│ │ ├── hr_9506_american_leadership.tex # New Statute Adapting FDORA Sec 3209, Cures Act
│ │ ├── hr_9507_data_self_custody.tex # Revision of HHS HIPAA, ONC Cures Rule
│ │ └── back_matter.tex
│ ├── research/ # Archived generation scripts
│ │ ├── v1.9.1/
│ │ │ ├── generate_pdf.py # reportlab + Pillow generator
│ │ │ ├── paper/README # Paper access (Drive link)
│ │ │ └── images/README.md # Image access (Drive link)
│ │ └── v1.9.0/
│ │ ├── README.md
│ │ ├── generate_illustrations.py
│ │ ├── generate_pdf.py
│ │ ├── paper/README
│ │ ├── svg/README.md
│ │ ├── pdf/README.md
│ │ └── png/README.md
│ └── prompts/
│ └── prompts.md # Development prompts archive
│
├── competitions/ # ★ Competitions and Instruction Sets
│ ├── instructions/ # ★ Glioblastoma Trial Instructions (v3.9.0)
│ │ ├── README.md # Top-level orientation and table of contents
│ │ ├── glioblastoma_context.md # Patient PAT-GBM-0001 + 5-phase procedure timeline
│ │ ├── robot_specification.md # Medtronic ROSA ONE Brain v3.0 + 50-channel sensors
│ │ ├── chunking_strategy.md # 3-layer chunking to keep future LLM in memory
│ │ ├── file_format_conventions.md # .md, .pdf, .json, .jsonl, .yaml, .toml, .parquet, etc.
│ │ ├── ascii_diagram_guide.md # ASCII + Mermaid templates that replace SVG
│ │ ├── runtime_environments.md # MacOS, Windows, Linux, Docker, server recipes
│ │ ├── competition_protocol.md # Prior versions, competitor robots, hybrid teams
│ │ ├── ci_compliance_checklist.md # Pre-commit ruff and yamllint checklist
│ │ ├── pr_workflow.md # 7-commit single-PR pattern
│ │ ├── commit_01_project_overview.md # Future Commit 1: 7 files
│ │ ├── commit_02_sensor_specifications.md # Future Commit 2: 8 files
│ │ ├── commit_03_xyz_mapping.md # Future Commit 3: 9 files
│ │ ├── commit_04_iteration_design.md # Future Commit 4: 10 files (incl. Cargo.toml)
│ │ ├── commit_05_comparison_competition.md # Future Commit 5: 12 files + release snapshot
│ │ ├── commit_06_error_fixes.md # Future Commit 6: 7-check pre-commit error scan
│ │ ├── commit_07_repository_updates.md # Future Commit 7: parent README, releases.md, CHANGELOG.md
│ │ └── one_minute_variant/ # ★ Glioblastoma 1-Minute Variant Instructions (v3.9.1)
│ │ ├── README.md # 1-min variant orientation and inheritance map from v3.9.0
│ │ ├── glioblastoma_context_1min.md # 4-phase 60-second procedure timeline (pre-op precomputed)
│ │ ├── robot_specification_neurospeed.md # Medtronic NeuroSpeed 1.0 (2030) 4 arms x 7 DOF
│ │ ├── sensor_specification_10khz.md # Mixed 1 kHz / 10 kHz force, 200 channels (50/arm x 4)
│ │ ├── multi_arm_coordination.md # 1 kHz heartbeat, 5 ms e-stop, 12 N cumulative cap
│ │ ├── file_size_pyramid_1min.md # Layer 4 pyramid: 510 KB committed + 26 MB Zenodo per iter
│ │ ├── commit_01_overview_1min.md # Future Commit 1: 8 files (vs parent 7)
│ │ ├── commit_02_sensors_1min.md # Future Commit 2: 9 files (vs parent 8)
│ │ ├── commit_03_xyz_4arm.md # Future Commit 3: 11 files (vs parent 9, adds heartbeat)
│ │ ├── commit_04_iterations_1min.md # Future Commit 4: 14 files (16 iter sweep at 1 min)
│ │ ├── commit_05_competition_1min.md # Future Commit 5: 13 files + v3.9.1 snapshot + Zenodo patch
│ │ └── zenodo_archive_protocol.md # 416 MB L0 raw deposition, DOI assignment, SHA-256 manifest
│ ├── inputs/ # Reference papers and Kaggle competition source files
│ │ ├── paper-a/ # CodeClash paper chunked into 10 markdown files
│ │ ├── paper-b/ # FAERS paper chunked into 10 markdown files
│ │ └── site-1/ # Orbit Wars Kaggle competition chunked into 3 files
│ ├── data/ # Future generated competition data
│ └── paper/ # Future generated competition paper
│
├── digital-twins/
│ ├── README.md
│ ├── patient-modeling/
│ │ ├── README.md
│ │ └── tumor_twin_pipeline.py
│ ├── treatment-simulation/
│ │ ├── README.md
│ │ └── treatment_simulator.py
│ ├── clinical-integration/
│ │ ├── README.md
│ │ └── clinical_dt_interface.py
│ └── examples-twins/
│ ├── README.md
│ ├── 01_realtime_dt_synchronization.py
│ ├── 02_multi_organ_toxicity_twin.py
│ ├── 03_adaptive_radiation_therapy_dt.py
│ ├── 04_tumor_microenvironment_immunotherapy_dt.py
│ ├── 05_virtual_trial_cohort_dt.py
│ └── 06_dt_validation_verification.py
│
├── examples/
│ ├── README.md
│ ├── 01_surgical_robot_training.py
│ ├── 02_digital_twin_surgical_planning.py
│ ├── 03_cross_framework_validation.py
│ ├── 04_agentic_clinical_workflow.py
│ └── 05_treatment_response_prediction.py
│
├── examples-new/
│ ├── README.md
│ ├── 01_realtime_safety_monitoring.py
│ ├── 02_sensor_fusion_intraoperative.py
│ ├── 03_ros2_surgical_deployment.py
│ ├── 04_hand_eye_calibration_registration.py
│ ├── 05_shared_autonomy_teleoperation.py
│ └── 06_robotic_sample_handling.py
│
├── q1-2026-standards/
│ ├── README.md
│ ├── objective-1-bidirectional-conversion/
│ │ ├── isaac_to_mujoco_pipeline.py
│ │ ├── mujoco_to_isaac_pipeline.py
│ │ └── physics_equivalence_tests.py
│ ├── objective-2-robot-model-repository/
│ │ ├── model_registry.yaml
│ │ └── model_validator.py
│ ├── objective-3-validation-benchmark/
│ │ └── benchmark_runner.py
│ └── implementation-guide/
│ ├── timeline.md
│ └── compliance_checklist.md
│
├── unification/
│ ├── README.md
│ ├── simulation_physics/
│ │ ├── challenges.md
│ │ ├── opportunities.md
│ │ ├── isaac_mujoco_bridge.py
│ │ ├── urdf_sdf_mjcf_converter.py
│ │ └── physics_parameter_mapping.yaml
│ ├── agentic_generative_ai/
│ │ ├── challenges.md
│ │ ├── opportunities.md
│ │ └── unified_agent_interface.py
│ ├── surgical_robotics/
│ │ ├── challenges.md
│ │ └── opportunities.md
│ ├── cross_platform_tools/
│ │ ├── framework_detector.py
│ │ └── validation_suite.py
│ ├── usl/ # ★ Unification Standard Level
│ │ ├── README.md # USL standard overview
│ │ ├── prompts.md # Development prompts archive
│ │ ├── paper/ # USL Paper (v1.8.0)
│ │ │ ├── Unification Standard Level for Physical AI Oncology Trials.pdf
│ │ │ ├── Latex Source Code.zip # .tex, .sty, .bib, README
│ │ │ ├── usl_oncology_trials.tex
│ │ │ ├── usl_oncology_trials_chunk/ # Chunked into 2 files (v2.9.1)
│ │ │ │ └── README.md # Reconstruction instructions
│ │ │ ├── usl-oncology.sty
│ │ │ ├── references.bib
│ │ │ └── README
│ │ ├── humanoids/ # ★ Humanoid Robots (v1.6.0)
│ │ │ ├── README.md # Humanoid evaluations & diagrams
│ │ │ ├── usl_humanoid_scoring.py
│ │ │ ├── boston_dynamics_atlas/
│ │ │ ├── tesla_optimus/
│ │ │ └── agility_digit/
│ │ ├── surgical/ # Surgical Robots (v1.5.0)
│ │ │ ├── README.md # Surgical evaluations & diagrams
│ │ │ ├── usl_surgical_scoring.py
│ │ │ ├── intuitive_davinci/
│ │ │ ├── medtronic_hugo/
│ │ │ └── cmr_versius/
│ │ └── cobots/ # Cobots (v1.4.0)
│ │ ├── README.md # Cobot evaluations & diagrams
│ │ ├── usl_scoring_framework.py
│ │ ├── franka_panda/
│ │ ├── kinova_gen3/
│ │ └── ufactory_xarm7/
│ ├── standards_protocols/
│ └── integration_workflows/
│
├── generative-ai/ # VLA models, diffusion policies, synthetic data
│ ├── strengths.md
│ ├── limitations.md
│ └── results.md
├── agentic-ai/ # LLM-based robot control, multi-agent systems
│ ├── README.md
│ ├── strengths.md
│ ├── limitations.md
│ ├── results.md
│ └── examples-agentic-ai/
│ ├── 01_mcp_clinical_robotics_server.py
│ ├── 02_react_procedure_planner.py
│ ├── 03_realtime_adaptive_treatment_agent.py
│ ├── 04_autonomous_simulation_orchestrator.py
│ ├── 05_safety_constrained_agent_executor.py
│ └── 06_protocol_rag_compliance_agent.py
├── reinforcement-learning/ # RL for surgical autonomy, sim2real transfer
│ ├── strengths.md
│ ├── limitations.md
│ └── results.md
├── self-supervised-learning/ # Contrastive learning, foundation models
│ ├── strengths.md
│ ├── limitations.md
│ └── results.md
├── supervised-learning/ # Segmentation, detection, classification
│ ├── strengths.md
│ ├── limitations.md
│ └── results.md
│
├── frameworks/
│ ├── nvidia-isaac/ # Isaac Sim, Isaac Lab, Isaac for Healthcare
│ │ └── INTEGRATION.md
│ ├── mujoco/ # MuJoCo, MJX, MuJoCo Playground
│ │ └── INTEGRATION.md
│ ├── gazebo/ # Gazebo Ionic, ROS 2 integration
│ │ └── INTEGRATION.md
│ └── pybullet/ # PyBullet medical simulation
│ └── INTEGRATION.md
│
├── privacy/
│ ├── README.md
│ ├── phi-pii-management/
│ │ ├── README.md
│ │ └── phi_detector.py
│ ├── de-identification/
│ │ ├── README.md
│ │ └── deidentification_pipeline.py
│ ├── access-control/
│ │ ├── README.md
│ │ └── access_control_manager.py
│ ├── breach-response/
│ │ ├── README.md
│ │ └── breach_response_protocol.py
│ └── dua-templates/
│ ├── README.md
│ └── dua_generator.py
│
├── regulatory/
│ ├── README.md
│ ├── Adaption-21-CFR-Part-312/ # ★ Physical AI 21 CFR Part 312 Adaptation (v2.5.0)
│ │ └── source/
│ │ ├── Physical_AI_21_CFR_Part_312.tex # LaTeX source (94 pages compiled)
│ │ ├── Physical_AI_21_CFR_Part_312_chunk/ # Chunked into 5 files (v2.9.1)
│ │ │ └── README.md # Reconstruction instructions
│ │ ├── Physical_AI_21_CFR_Part_312.sty # Custom style package
│ │ ├── Physical_AI_21_CFR_Part_312.bib # Bibliography (42 references)
│ │ ├── Physical_AI_21_CFR_Part_312.pdf # Compiled PDF
│ │ ├── Physical_AI_21_CFR_Part_312.zip # Source archive
│ │ └── prompts.md # Development prompts archive
│ ├── Adaption-21-CFR-Part-50/ # ★ Physical AI 21 CFR Part 50 Adaptation (v2.4.0)
│ │ └── source/
│ │ ├── Physical_AI_21_CFR_Part_50.tex # LaTeX source (37 pages compiled)
│ │ ├── Physical_AI_21_CFR_Part_50_chunk/ # Chunked into 3 files (v2.9.1)
│ │ │ └── README.md # Reconstruction instructions
│ │ ├── Physical_AI_21_CFR_Part_50.sty # Custom style package
│ │ ├── Physical_AI_21_CFR_Part_50.bib # Bibliography (19 references)
│ │ ├── Physical_AI_21_CFR_Part_50.pdf # Compiled PDF
│ │ ├── Physical_AI_21_CFR_Part_50.zip # Source archive
│ │ ├── prompts.md # Development prompts archive (v2.4.0)
│ │ └── README.md
│ ├── adaption-ich-e6r3/ # ★ Physical AI Unification Guidance (v2.2.0)
│ │ ├── prompts.md # Development prompts archive
│ │ └── source/
│ │ ├── main.tex # LaTeX source (Sections 1-4, Appendices, Glossary)
│ │ ├── main_chunk/ # Chunked into 4 files (v2.9.1)
│ │ │ └── README.md # Reconstruction instructions
│ │ ├── ich_guideline_style.sty
│ │ ├── references.bib
│ │ ├── compiled.pdf
│ │ └── README.md
│ ├── fda-compliance/
│ │ ├── README.md
│ │ └── fda_submission_tracker.py
│ ├── irb-management/
│ │ ├── README.md
│ │ └── irb_protocol_manager.py
│ ├── ich-gcp/
│ │ ├── README.md
│ │ └── gcp_compliance_checker.py
│ └── regulatory-intelligence/
│ ├── README.md
│ └── regulatory_tracker.py
│
├── regulatory-submit/ # FDA Submission Automation (v1.0.0)
│ ├── README.md
│ ├── presub_generator.py # Pre-Submission (Q-Sub) package generation
│ ├── pccp_engine.py # PCCP document authoring
│ ├── classification_advisor.py # 510(k)/De Novo/PMA pathway classification
│ ├── iec62304_generator.py # IEC 62304 lifecycle documentation
│ ├── clinical_evidence.py # Clinical evidence reports
│ ├── audit_trail.py # 21 CFR Part 11 audit trails
│ └── examples-regulatory-submit/ # 6 submission workflow examples
│
├── federation/
│ ├── README.md
│ ├── federated_coordinator.py
│ ├── differential_privacy.py
│ ├── secure_aggregation.py
│ ├── site_enrollment.py
│ ├── data_harmonization.py
│ ├── consortium_reporting.py
│ ├── privacy_analytics.py
│ └── examples-federation/
│ ├── README.md
│ ├── 01_basic_two_site.py
│ ├── 02_differential_privacy.py
│ ├── 03_secure_aggregation.py
│ ├── 04_enrollment_sync.py
│ ├── 05_data_harmonization.py
│ └── 06_full_consortium.py
│
├── tools/
│ ├── README.md
│ ├── dicom-inspector/
│ │ └── dicom_inspector.py
│ ├── dose-calculator/
│ │ └── dose_calculator.py
│ ├── trial-site-monitor/
│ │ └── trial_site_monitor.py
│ ├── sim-job-runner/
│ │ └── sim_job_runner.py
│ └── deployment-readiness/
│ └── deployment_readiness.py
│
├── images/
│ ├── README.md
│ ├── prompts/ # Human-authored + AI-recommended prompts
│ │ ├── plan.md
│ │ ├── 1st.md
│ │ ├── 2nd.md
│ │ └── 3rd.md
│ ├── interactive/ # Python visualization scripts
│ │ ├── 1st/ # 10 scripts (architecture, clinical)
│ │ ├── 2nd/ # 10 scripts (AI/ML benchmarks)
│ │ └── 3rd/ # 10 scripts (regulatory, privacy)
│ └── png/ # Static PNG exports (1920×1080 @2x)
│ ├── 1st/ # 20 PNGs (10 light + 10 dark)
│ ├── 2nd/ # 20 PNGs (10 light + 10 dark)
│ └── 3rd/ # 20 PNGs (10 light + 10 dark)
│
├── national-platform/ # ★ National Platform for Physical AI Oncology Trials
│ ├── RESEARCH-A # Federal regulatory research (plain text)
│ ├── RESEARCH-A-CHUNK/ # Chunked into 2 files (v2.9.1)
│ │ └── README.md # Reconstruction instructions
│ ├── RESEARCH-B # State/federal comparative research (plain text)
│ ├── RESEARCH-B-CHUNK/ # Chunked into 2 files (v2.9.1)
│ │ └── README.md # Reconstruction instructions
│ ├── 21cfr312_adapt/ # 21 CFR Part 312 adaptation chunks (5 files)
│ ├── 21cfr50_adapt/ # 21 CFR Part 50 adaptation chunks (3 files)
│ ├── ich_e6r3_adapt/ # ICH E6(R3) adaptation chunks (4 files)
│ ├── federated_learning/ # Federated learning paper chunks (4 files)
│ ├── national_mcp/ # National MCP servers paper chunks (4 files)
│ ├── new_trial_psl/ # Trial site PSL documentation chunks (11 files)
│ ├── patient_journey/ # Patient journey paper chunks (3 files)
│ ├── patient_robot/ # Patient robot instructions chunks (2 files)
│ ├── usl_standard/ # USL standard paper chunks (2 files)
│ ├── research_a/ # Research A analysis chunks (2 files)
│ ├── research_b/ # Research B analysis chunks (2 files)
│ ├── paper_template/ # Original Groningen LaTeX template
│ ├── new_template/ # National Platform LaTeX template (v2.9.2)
│ │ ├── main.tex # Template entry point (16 sections)
│ │ ├── page_styles.tex # Page styles with attribution
│ │ ├── references.bib # Bibliography (35 sources)
│ │ ├── README.md # Template documentation
│ │ └── sections/ # 20 section .tex files
│ └── new_paper/ # ★ Compiled National Platform Paper (v3.0.0)
│ ├── main.tex # Main document (191 pages)
│ ├── main.pdf # Compiled PDF
│ ├── page_styles.tex # Page styles
│ ├── references.bib # Bibliography (34 sources, clickable URLs/DOIs)
│ ├── latex_source.zip # Complete LaTeX source archive
│ ├── README.md # Paper documentation
│ └── sections/ # 21 section .tex files
│
├── configs/
│ └── training_config.yaml
│
└── scripts/
└── verify_installation.py
| Framework | Version | Last Update | Use Case | Unification Status |
|---|---|---|---|---|
| NVIDIA Isaac Lab | 2.3.1 | Dec 2024 | GPU-accelerated robot training | ✓ Bridge available |
| NVIDIA Isaac Sim | 5.0.0 | Jan 2026 | High-fidelity physics simulation | ✓ Bridge available |
| Newton Physics Engine | Beta | Jan 2026 | GPU physics (NVIDIA/DeepMind/Disney) | ✓ Isaac Lab integrated |
| MuJoCo | 3.4.0 | Dec 2024 | Precision physics simulation | ✓ Bridge available |
| MuJoCo Warp | Beta | Jan 2026 | GPU-optimized MuJoCo (NVIDIA) | ✓ Bridge available |
| Gazebo Sim (Jetty) | 10.0.0 | Oct 2024 | ROS 2 integrated simulation | ◐ In progress |
| PyBullet | 3.2.5 | Apr 2023 | Rapid prototyping | ✓ Bridge available |
| Framework | Stars | Last Update | Use Case | Unification Status |
|---|---|---|---|---|
| NVIDIA GR00T N1.6 | - | Jan 2026 | Humanoid robot foundation model | ✓ Adapter available |
| NVIDIA Cosmos Predict 2.5 | - | Jan 2026 | World foundation model, synthetic data | ✓ Native support |
| NVIDIA Cosmos Reason 2 | - | Jan 2026 | Reasoning VLM for physical AI | ✓ Native support |
| CrewAI | 100K+ | Jan 2026 | Multi-agent orchestration (v1.6.1) | ✓ Unified interface |
| LangChain/LangGraph | 95K+ | Jan 2026 | LLM-robot integration (v1.1.0) | ✓ Unified interface |
| Model Context Protocol | - | Dec 2025 | Agent-tool communication (AAIF/Linux Foundation) | ✓ Native support |
| MONAI Multimodal | - | Jan 2026 | Medical imaging + agentic AI | ✓ Integrated |
| Framework | Institution | Last Update | Use Case | Unification Status |
|---|---|---|---|---|
| ORBIT-Surgical | Stanford/JHU | Dec 2024 | Surgical task benchmarking | ✓ Primary benchmark |
| dVRK 2.4.0 | JHU | Jan 2026 | da Vinci research platform (ROS 2 Jazzy) | ✓ Bridge available |
| dVRK-Si | JHU | 2025 | Next-gen da Vinci Si/S support | ✓ Bridge available |
| SurgicalGym | - | 2025 | GPU-based surgical RL | ◐ In progress |
| Isaac Lab-Arena | NVIDIA | Jan 2026 | Large-scale policy evaluation | ✓ Benchmark integration |
The new digital-twins/ directory provides comprehensive tools for creating and using patient-specific digital twins in oncology clinical trials.
| Capability | Framework | Clinical Application |
|---|---|---|
| Tumor growth modeling | TumorTwin | Patient-specific progression prediction |
| Treatment simulation | Custom PK/PD | Response prediction before treatment |
| Surgical planning | Isaac Sim integration | Virtual surgery rehearsal |
| Clinical integration | FHIR/DICOM | Hospital system connectivity |
python>=3.10
torch>=2.5.0
numpy>=1.24.0
scipy>=1.11.0
# NVIDIA Isaac (requires NVIDIA GPU)
isaacsim>=5.0.0
isaaclab>=2.3.0
# MuJoCo
mujoco>=3.4.0
mujoco-mjx>=3.4.0 # JAX backend
# ROS 2 (Kilted Kaiju or Jazzy)
ros-jazzy-desktop # or ros-kilted-desktop
# Agentic AI
langchain>=1.0.0
langgraph>=1.0.0
crewai>=1.0.0
All referenced repositories have been updated within October 2025 - March 2026:
| Repository | Purpose | Last Commit |
|---|---|---|
| isaac-sim/IsaacLab | Robot learning framework (v2.3.1) | Dec 2024 |
| newton-physics/newton | GPU physics engine (Linux Foundation) | Jan 2026 |
| google-deepmind/mujoco | Physics simulation (v3.4.0) | Dec 2024 |
| google-deepmind/mujoco_warp | GPU-optimized MuJoCo | Jan 2026 |
| orbit-surgical/orbit-surgical | Surgical simulation | Sep 2024 |
| jhu-dvrk/sawIntuitiveResearchKit | dVRK platform (v2.4.0) | Jan 2026 |
| NVIDIA/Isaac-GR00T | GR00T N1.6 foundation model | Jan 2026 |
| RobotecAI/rai | ROS 2 agentic framework | Active |
| crewAIInc/crewAI | Multi-agent orchestration (v1.6.1) | Jan 2026 |
| langchain-ai/langgraph | Durable agent framework (v1.1.0) | Jan 2026 |
| modelcontextprotocol | MCP specification (AAIF) | Jan 2026 |
| Project-MONAI/MONAI | Medical imaging AI | Jan 2026 |
| SCAI-Lab/ros4healthcare | Healthcare robotics | 2025 |
| bulletphysics/bullet3 | Physics engine (v3.2.5) | Apr 2023 |
| OncologyModelingGroup/TumorTwin | Patient-specific cancer DTs | 2025 |
| surgical-robotics-ai | Surgical robotics ML | Active |
| SamuelSchmidgall/SurgicalGym | GPU surgical simulation | Active |
| med-air/SurRoL | dVRK-compatible RL platform | 2025 |
The unification framework supports collaboration across institutions:
| Organization Type | Contribution Area | Integration Point |
|---|---|---|
| Academic Labs | Algorithms, benchmarks | ORBIT-Surgical, skill library |
| Industry R&D | Hardware, deployment | ros2_surgical, safety validation |
| Healthcare Systems | Clinical validation | Multi-site coordination |
| Regulatory Bodies | Compliance standards | IEC 62304 documentation |
| Privacy Officers | PHI/PII management, de-identification | privacy/ framework tools |
| Regulatory Affairs | FDA/IRB/ICH-GCP compliance | regulatory/ framework tools |
See unification/README.md for the complete cooperation model.
If you use this repository in your research, please cite using BibTeX:
@software{kawchak2026physicalai,
author = {Kawchak, Kevin},
title = {Physical AI for Oncology Clinical Trials},
version = {3.4.1},
year = {2026},
publisher = {GitHub},
url = {https://github.com/kevinkawchak/physical-ai-oncology-trials}
}MIT License - See LICENSE for details.