Humanoid Simulation

Status Quo

Existing oil rigs, particularly brownfield rigs, were designed around human operators with controls, gauges, and equipment layouts optimized for manual operation. These rigs cannot be easily redesigned, and retrofitting them with smart controls and digital gauges would require significant capital investment and operational disruption.

Humanoid and quadruped robots offer an alternative approach. Rather than requiring equipment modifications, robots can adapt to existing infrastructure. The humanoid form factor is particularly advantageous because controls and interfaces were already built for human operators. Additionally, extensive human video data enables transfer learning and accelerated model training, while general-purpose humanoids could evolve into a generalized integrator platform across different customer sites and equipment configurations.

NOV customers have already expressed interest in robotic solutions, including for geothermal use cases and mud systems automation, creating a clear market opportunity for this technology.

Goal

This project explores the viability and strategic value of humanoid and mobile robots for oil rig operations, aiming to reduce operational costs, increase safety, and address persistent workforce shortages. Our goal is to define, simulate, and validate robot tasks for offshore oil rigs that minimize human operator requirements, ultimately positioning NOV as the industry leader in practical humanoid applications for drilling and geothermal operations.

We'll use an existing 3D scanned offshore oil rig as a proving ground. This allows us to test various types of robots and their performance in realistic conditions. We use NVIDIA IsaacSim as our simulation platform to test physically realistic behavior of real robots in this virtual environment before any hardware investment.

Since the degree of automation on the drill floor is already high on newer rigs, the focus will be on auxiliary systems, such as mud handling. These systems will need to be modeled, either from 3D scans, CAD drawings, or publicly available assets.

Key Challenges

Several challenges must be addressed for successful deployment:

• Form Factor Trade-offs: Not all tasks may benefit from humanoid form. Some scenarios may work better with specialized grippers rather than human-like hands.
• Harsh Environment Requirements: Explosion-proof, harsh-environment-rated hardware is not yet readily available for humanoid platforms.
• Complex Physical Interactions: Robots must handle mud, valves, gauge reading, screen shaker maintenance, stairs, and repetitive workflows in challenging conditions.
• Limited Connectivity: Rigs typically lack Wi-Fi infrastructure, requiring reliance on edge compute or satellite connectivity like Starlink.

Proposed Solution

Environment

Our virtual environment builds on existing 3D scans and open source assets of oil rigs. We enrich this base environment with high-fidelity 3D assets to create a more flexible simulation environment that supports complex manipulation tasks.

These assets include pallets with equipment, wiring harnesses, hydraulic hoses and junctions, as well as specialized drilling equipment like the iron roughneck and topdrive. We also incorporate existing robots, containers, doors, and control screens to create a realistic operational environment for testing, that can be physically modeled and manipulated.

NVIDIA IsaacSim

NVIDIA IsaacSim provides a physics-based robotics simulation platform built on OpenUSD.

This combination allows us to quickly model realistic scenes with modular, reusable assets. We can describe complex environments, equipment, and robots using OpenUSD's compositional architecture, enabling rapid iteration and collaborative development.

The platform includes physically correct simulation of sensors crucial for autonomous operation. This includes lidar, cameras, depth sensors, and IMUs, all simulated with realistic noise models and physical constraints Robots can test their perception and navigation algorithms in conditions that closely match real-world physics.

NVIDIA GR00T

NVIDIA GR00T is a foundation model designed specifically for humanoid robots. It provides a pre-trained baseline for robot learning that can be fine-tuned for specific tasks and environments. By leveraging GR00T in our simulation environment, we can accelerate the development of robot behaviors and reduce the training data requirements for deployment.

The model is designed to work seamlessly with IsaacSim, allowing us to generate synthetic training data in simulation and transfer learned behaviors to physical robots. This approach significantly reduces the time and cost associated with training robots for complex manipulation and navigation tasks in oil rig environments.

Robots

Using existing assets in IsaacSim, the team can quickly bring realistic robots into the virtual environment. To evaluate performance of different solutions, we will model several approaches:

Focus: Humanoid	Optional: Quadruped	Optional: Manipulator
Humanoid with hand end-effectors	Quadruped with manipulator mounted on top	Simple cobot manipulator, placed near control surfaces
Most flexible solution	Move freely through environment	Simple, cheap, reliable
Least proven in industrial settings	Trouble navigating tight spaces	Not flexible, fixed mounting location

Each robot type offers different trade-offs for oil rig automation. Simple manipulators could be mounted next to control surfaces, offering reliability and low cost but limited to fixed locations. Quadrupeds provide better mobility across the rig but struggle in confined areas and lack the dexterity of humanoid solutions. Humanoids represent the most flexible option, capable of operating existing human-designed controls and navigating tight spaces. However, they remain the least proven technology in harsh industrial environments. Testing these options in simulation will help validate whether new-generation humanoids can effectively operate on existing oil rigs.

Depending on the prioritized task list, the team may also decide to focus only on humanoids and their different form factors and end effectors.

Candidate Tasks

The simulation will focus on validating high-value automation scenarios identified through customer feedback and operational analysis:

Approach

We propose an iterative approach that first identifies the most likely candidates for robotic automation, implements simple test scenarios, and validates that a high degree of autonomous operations is feasible.

Task Identification & Assessment

Define and prioritize a list of humanoid-suitable tasks.

Deliverables:

• Task Catalogue including detailed descriptions, required capabilities, environmental constraints, safety factors, and industrial standards.
• Feasibility assessment per task (technical difficulty, cost impact, potential ROI).
• Ranking & selection of 5–10 high-value humanoid use cases.

Initial Candidate Tasks:

• Gauge reading
• Valve turning
• Shaker screen removal and latch manipulation
• Mud sack handling and mud systems interactions
• Crane interaction / repeating human workflows
• Light maintenance tasks
• Geothermal operation tasks

The team will conduct structured workshops and expert interviews with NOV SMEs, including David, Ali, and Brad Wood.

Virtual Environment & Robotics Assets

Import scanned data from one of NOV's R&D facility into IsaacSim. Recreate the auxiliary systems on an offshore rig, ensuring all critical components are accurately represented. This includes modeling the inputs and outputs used for automation scenarios, such as control panels, valves, and monitoring equipment. Start testing with quadruped and manipulator robots to establish baseline performance for simpler automation tasks.

Deliverables:

• Digital twin environment of one of NOV's offshore R&D facilities in IsaacSim
• Asset library of modular rig components (control panels, valves, screens, equipment)
• Robot models configured with realistic kinematics and sensor suites
• Initial baseline tests documenting robot reachability and basic manipulation capabilities
• Technical documentation of environment setup and asset integration workflows

Building Blocks for Scenarios

Focus moves to implementing and validating automation scenarios. Begins with a basic drilling workflow in simulation to validate the environment setup. Next, the team develops more complex scenarios that require path planning and task execution based on behavior trees. For each scenario, KPIs including execution speed, failure rate, and robot uptime versus downtime are calculated. These metrics help understand which robot types perform best for specific tasks.

Deliverables:

• Simulation environment with at least 5–10 validated automation workflows
• Task performance analysis including reachability, cycle time, precision, and environmental hazards
• Behavior tree implementations for path planning and task execution
• KPI dashboard showing execution speed, failure rates, and robot uptime metrics
• Data assets for potential training pipelines and future model development
• Video demonstrations of key scenarios for stakeholder communication

Running Scenario Simulations & Calculating ROI

In the final phase, the team implements a majority of high-value automation scenarios identified during the discovery process. Run different robot types through these scenarios to compare their performance in realistic operational conditions. By calculating overall equipment efficiency and comparing it with baseline human crew performance, the potential impact of robotic automation is quantified. This analysis culminates in a comprehensive ROI calculation that considers initial investment, operational costs, and productivity gains.

Deliverables:

• Comprehensive ROI analysis comparing robotic automation to human crew baseline
• Overall equipment efficiency (OEE) calculations for each robot type and scenario
• Platform comparison matrix rating robot solutions by cost, suitability, and performance
• Partner landscape and rating of hardware manufacturers, gripper specialists, and integrators
• Integration concept for NOV equipment and robotics workflows
• Pilot test plan for physical validation at NOV test center
• Executive summary with strategic recommendations for NOV's robotics roadmap

Expected Outcomes

This project will deliver clear, actionable insights for NOV's robotics strategy:

Technical Validation: Simulation-based validation of feasibility before any hardware investment, reducing risk and accelerating decision-making.

Strategic Positioning: A foundation for becoming the industry leader in practical humanoid applications for drilling and geothermal operations.

Clear Task Definition: Comprehensive documentation of high-impact humanoid workflows with feasibility assessments and prioritization.

Partner Ecosystem: Identification of hardware manufacturers, gripper specialists, and explosion-proof robotics integrators to support future deployment.

Customer Engagement Assets: Video material and simulation demonstrations that NOV can use to engage customers and demonstrate thought leadership.

Multi-Year Roadmap: A foundation for continued development that can be funded jointly by NOV and participating operators.

Work packages

Roles & Cost

Rate Card