kamal01k / vehicle-simulation-toolkit

Unity Vehicle Simulation Environment, Driving Scenario Visualization Tool & Driving Scenario Generation Tool

Vehicle Simulation Toolkit

This toolkit provides building blocks for the development of simulated driving scenarios. Three components included are:

1. Unity Vehicle Simulation Environment: Construct 3D environments with dynamic objects such as vehicles and pedestrians.
2. Driving Scenario Vehicle Visualization Tool: Visualize driving scenarios in a 3D world from simulated or recorded data.
3. Driving Scenario Generation Tool: Prototype scenario generation tool for crowdsourcing driving environments.

Unity Vehicle Simulation Environment

Creating realistic vehicle simulation environments fundamentally requires many different dynamic objects that interact with each other. Common dynamic objects include environmental signals, adjacent vehicles, and pedestrians. The Unity Vehicle Simulation Environment provides the scripts needed to create realistic driving scenarios using Unity. It includes dynamic objects and road generation scripts.

Note: This package includes scripts which are dependent on additional external libraries. The Unity Standard Assets is not included in the project. Please import them from the Unity Assets Store to fix dependencies.

Driving Scenario Vehicle Visualization Tool

This tool provides a means to visualize driving scenarios with from simulated or recorded data. It also captures images/videos from the scene at from any position desired. The tool also provides opportunities to explore computer vision algorithms by providing depth images and pixel-level labeled data sets.

1. Provide visual feedback about a driving scenario
2. Support research for creating large data sets of driving scenarios
3. Assess performance of vision algorithms

Driving Scenario Vehicle Visualization API

This section provides details for the vehicle visualization API and explains how to provide data to the system in the appropriate format. All files are in JSON format and must contain the appropriate fields for the system. Files are placed in separate folders depending on the vehicle described. Check out the "Scenario Example" folder for a template.

Dynamic Agents (Vehicles and Pedestrians)

Agents use high quality detailed 3D models. Each agent in the scene must be described with a unique JSON file located in the folder. Vehicles must be placed in the vehicle folder and pedestrians must be placed in the pedestrian folder. The format is the following:

{ "time": Array.of(float), "x": Array.of(float), "y": Array.of(float), "z": Array.of(float), "phi": Array.of(float), "psi": Array.of(float), "theta": Array.of(float) }

Environment

Environment can contain a variety of features: Roads, signs, buildings. Roads can be made by using lanes coordinates or a custom road builder.

{ "x": Array.of(float), "y": Array.of(float), "z": Array.of(float), "LaneWidth": float }

Coordinate System

The vehicle visualization tool uses Unity's left handed coordinate system. All units are in meters and angles are in degrees.

Interacting with the Scene

C - Change the position of the camera. There are three positions available - first person, third person, and third person orbit. T - Toggle the target of the camera to another vehicle in the scene. R - Reset the scene. S - Stop the scene. P - Play the scene if stopped. Q - Quit.

Run the executable from the command prompt with a string containing the location of your scenario.

VehicleVisualizationTool.exe "...\Scenario Library\Scenario Example"

Driving Scenario Generation Tool

This is a prototype application to design and crowdsource driving scenarios to test autonomous vehicles in novel situations. It provides a simple user interface to add vehicles and pedestrians onto a virtual environment and add kinematic behaviors for each agent in the scene. The sensing of each agent is simulated (visual and depth) and can be used to test vehicle controllers. Future iterations of the application may include the ability to add more realistic surrounding driving environments and more detailed behavior models for surrounding dynamic agents.