Immersiverse

Building Intelligence for the Physical World

We build simulation and AI technologies that enable machines to perceive, learn, decide and operate in complex physical environments.

ENISA Certified

Recognised as an Empresa Emergente under Spain’s Startup Law

Our Projects

At ImmersiVERSE, we develop simulation-driven technologies for Physical AI, robotics, and autonomous systems, combining physics-based digital twins, realistic sensor simulation, synthetic data generation, AI training, reinforcement learning, and XR teleoperation. Our projects span intelligent robots, autonomous vehicles, UAVs and UGVs, industrial environments, and challenging real-world conditions such as rain, fog, smoke, wind, and night, enabling AI systems to be trained, tested, and validated in realistic virtual environments before deployment in the physical world.

Physics-Based Modelling

We build and research physics-based mathematical models that capture the complexity of real-world systems.

Sensor Simulation

Use mathematical models to reproduce realistic LiDAR, radar, and camera measurements, including noise, range, resolution, and sensor-specific behaviour

Weather Modelling

Model rain, fog, snow, and atmospheric effects mathematically to simulate their impact on sensor signals, detection probability, and measurement uncertainty.

Synthetic Data

Generate physically consistent sensor outputs and ground-truth data enabling scalable AI training, validation, and perception testing.

Vehicle Dynamics

Apply physics-based mathematical models for vehicle motion, including acceleration, steering, braking, tire forces, and 6-DOF dynamics.

We are building IVerseGym, a simulation-based training environment for robots and autonomous vehicles using Reinforcement Learning and XR teleoperation. It enables agents to learn complex behaviors through policies trained across diverse simulated environments, with extensive domain randomization across physics, sensors, weather, terrain, objects, and operating conditions. The goal is to accelerate sim-to-real transfer and develop robust AI policies that can perform reliably in the real world

Technology Stack

Physical AI

Physics-aware agents that perceive, decide, and act through realistic forces, constraints, and sensor feedback.

Digital Twins

High-fidelity virtual replicas of real-world assets or environments used for analysis, optimization, and continuous validation

Real-World Data Augmentation

Integration of Gaussian splats, sensor data, and trajectories from deployed systems to ground simulation in reality

Synthetic Data & Edge-Case Generation

Scalable creation of labeled data, including rare, unsafe, or hard-to-capture scenarios.

Reinforcement Learning

Systematic variation of physics, visuals, and parameters to improve robustness and generalization for policies

Perception Simulation

Emulating sensors like cameras, LiDAR, radar, or IMUs to train and test AI perception modules in different and challenging edge scenerios

Unprecedented speed and precision

Enabling the creation of precise dataset of millions of  images  in a single day, revolutionizing the pace of AI development

Rapid prototyping

Iterate and refine AI vision models, streamlining the development process for quicker deployment and optimal performance.

Highly cost effective

Reduces costs by enabling AI vision systems to train and optimize in a virtual environment, sparing expensive real-world data collection and hardware expense

Data Privacy

we ensure that your sensitive information remains protected without compromise. 

Safety

Ensures safety for AI vision systems by allowing them to learn and adapt in controlled virtual environments, minimizing the risk associated with real-world experimentation and potential harm

Scalabilty and usabilty

 our simulation-driven approach empowers AI vision systems to effortlessly adapt to evolving demands, ensuring seamless integration and robust performance at any scale