Skip to content

Synthex Simulation Engine

Simulate the Scenarios the Real World Cannot Easily Provide

Explore virtual environments, configurable conditions and repeatable scenarios for developing and evaluating physical AI systems.

Engine 02Platform VisionIn development — not generally available.
Virtual sceneIllustrative
Fog density
Sun elevation
Agent density
Sensor noise
Illustrative virtual road scene with a vehicle, a pedestrian and an ego vehicle whose camera and depth sensor rays cover the road, under fog, with example scenario parameters.

Capabilities

Control the conditions, not just the data

Planned capabilities for describing, configuring and capturing simulated scenarios.

  • Platform Vision

    Scenario Design

    Describe environments, actors, conditions and success criteria as structured, versioned scenarios.

  • Platform Vision

    Virtual Environment Configuration

    Assemble roads, fields or indoor spaces from configurable assets and layouts.

  • Platform Vision

    Weather and Lighting Variation

    Sweep rain, fog, sun angle and low light to study how perception degrades.

  • Platform Vision

    Object and Agent Behaviour

    Place vehicles, people, plants or objects and script how they move and interact.

  • Platform Vision

    Virtual Sensor Capture

    Model camera and depth sensors with configurable placement, resolution and noise.

  • Platform Vision

    Edge-Case Simulation

    Generate rare, hazardous or ambiguous situations on demand and repeat them exactly.

How to read status labels

Platform Vision
A planned capability that defines where the platform is heading. Not yet available.
Research Direction
Active exploration of methods and questions. No prototype or results yet; not a product.
Illustrative Demonstration
A visual or interactive explanation using sample content. No live system or real results behind it.
Validated Implementation
Implemented and verified against defined criteria. No capability on this site is presented at this level yet.

Interactive Scenario Concept

Configure a scenario

Choose an environment, conditions and complexity to see how a scenario could be described. This is an educational illustration: it updates a description and a drawing, and does not run a simulator.

Interactive Scenario Concept

Illustrative Demonstration
Environment
Conditions
Complexity
ILLUSTRATION · NOT A SIMULATION OUTPUT

urban.fog.moderate

Urban Road · Fog · Moderate

An urban road scene for testing pedestrian and vehicle perception, in fog conditions, with typical density with partial occlusion.

Virtual sensors
Front RGB camera, depth
Ground truth
Boxes, masks, depth
Visibility
Low, distance-dependent
Lighting
Diffuse
Layout
Signalised junction
Vehicles
10
Pedestrians
6
Occlusion
Partial

Educational illustration. Changing options updates this description and drawing only — no simulator is run and no data is generated. It is not connected to CARLA, Unreal Engine, Unity, AirSim or any other simulation environment.

Simulation pipeline

From scenario definition to model evaluation

The intended flow from a written scenario to data that can be used for training and evaluation. Simulation execution is shown conceptually; no simulation backend is integrated yet.

Execution is conceptual

No simulation backend is integrated or validated yet. Steps marked “Conceptual” describe intended behaviour.
  1. 01

    Scenario Definition

    State the situation, actors, conditions and what the system under test must handle.

  2. 02

    Environment Setup

    Load or assemble the virtual environment and its static layout.

  3. 03

    Object and Condition Configuration

    Place agents and objects; set weather, lighting and variation ranges.

  4. 04

    Simulation ExecutionConceptual

    Run the configured scenario variations in a simulation backend.

  5. 05

    Sensor/Data Capture

    Record virtual sensor outputs together with exact ground truth.

  6. 06

    Synthetic Dataset Output

    Hand captured samples and labels to the Data Engine for validation and packaging.

  7. 07

    Model Evaluation

    Test models in the Evaluation Engine on the same scenarios, repeatably.

Applications

Where controlled simulation helps

  • Autonomous Driving

    Difficult driving conditions and perception edge cases.

    Example scenarios

    • Pedestrians emerging from occlusion at dusk
    • Lane markings degraded by rain and glare
    • Unusual vehicles and road works layouts
  • Agriculture

    Crop variation, weed distribution and changing field conditions.

    Example scenarios

    • Weed density and species mixes across a season
    • Growth stages, soil types and residue cover
    • Low sun angles, shadows and wet foliage
  • Robotics

    Object interactions, visibility challenges and sensor variability.

    Example scenarios

    • Cluttered, partially occluded objects to grasp
    • Reflective and transparent surfaces
    • Camera placement and calibration drift

Technology approach

Building on established simulation technology

Rather than re-creating rendering and physics, Synthex is exploring how existing simulation platforms could sit behind a common scenario and capture layer. The technologies below are under evaluation; there is no partnership with, or endorsement by, their owners.

Synthex scenario layer (planned)

A backend-independent scenario format, parameter sweeps, capture configuration and dataset hand-off to the Data Engine.

Platform Vision
  • CARLA

    Under exploration

    Open-source simulator for autonomous driving research, built on Unreal Engine, with configurable sensors, maps and traffic.

  • Unreal Engine

    Under exploration

    Real-time 3D engine from Epic Games, used for photorealistic rendering of large environments.

  • Unity

    Under exploration

    Real-time 3D development platform with tooling that can be used for synthetic data generation.

  • AirSim

    Under exploration

    Open-source simulator for drones and vehicles originally released by Microsoft; the original project is no longer actively developed.

CARLA, Unreal Engine, Unity and AirSim are trademarks or names of their respective owners. Mention here describes technologies under exploration and does not imply affiliation.

Next steps

Explore Controlled Environments for Real-World AI

If your models struggle with conditions that are hard to capture, we would like to hear which scenarios matter most to you.