Autopilot MATLAB toolbox (Automated Driving Toolbox) Introduction

Automatic driving fast in recent years, many companies want a share in this industry. Since 2017, MathWorks MATLAB company also hopes for the new autopilot toolbox, and have updated annually, simple to understand and summarize below the main functions of MATLAB toolbox under autopilot.

1 Overview

The MathWorks official gateway to introduce the main toolbox as follows:

Autopilot Toolbox ™ ADAS (driver assistance system) and autopilot system provides algorithms and tools for design, simulation and test mainly. Users can design and test the visual, radar, and a laser sensing system, sensor fusion, path planning and a vehicle controller. Providing visual tools: a sensor cover, aerial view detection and tracking, to display a video, and the laser radar map. Toolbox allows you to import and use of this high-definition real-time map data and OpenDRIVE® road network.

Use Ground Truth Labeler App, the user may be automatically labeled ground truth (true value), in order to train and evaluate the sensing algorithm; For a hardware fused ring (HIL) testing and sensing, sensor, the desktop emulation path planning and control logic may be generated and simulated driving scenarios. It can be simulated in the photorealistic three-dimensional environment camera, radar and lidar sensor output and an analog sensor for detecting an object in the lane boundaries and 2.5-D simulation environment.

Toolbox is a common ADAS autopilot and autopilot functions (including FCW, AEB, ACC, LKA parking and valet parking) to provide a reference application examples. The toolbox supports C / C ++ code can be generated for rapid prototyping and HIL testing, and supports the sensor fusion, track, path planning algorithms and application-level vehicle controller.

Official Photo link: https://www.mathworks.com/products/automated-driving.html

2 Function

2.1 true value tagging (Ground Truth Labeling)

2.1.1 Automatic tagging

Ground Truth Labeling App using interactive and automated ground truth labels, to facilitate target detection, segmentation and semantic scene classification.

2.1.2 Test Sensing Algorithm

By comparing the true value, the output of the sensing accuracy test.

2.2 driving scene simulation (Driving Scenario Simulation)

2.2.1 cuboid drive emulation (Cuboid Driving Simulation)

Generating a synthetic detected from the radar sensor and the camera model, and combine these driving scene is detected using the autopilot based simulator test cuboid algorithm. App using the driving scene designer defines the road network, road participant (cars, pedestrians, cyclists, obstacles and the like) and a sensor. The introduction of pre-built Euro NCAP test and OpenDRIVE road network.

2.2.2 virtual scene simulation engine driving (Unreal Engine Driving Scenario Simulation)

Unreal Engine® rendered 3D simulation environment using Epic Games® in the development, testing and visual driving algorithm.

2.3 calculated based on the perceptual visual and radar (Perception with Computer Vision and Lidar)

2.3.1 Vision System Design (Vision System Design)

Development of computer vision algorithms for the detection of vehicles, pedestrians, roads and classification.

2.3.2 Processing lidar (Lidar Processing)

Using a laser radar data to detect an obstacle and segment ground planes (to be analyzed).

2.4 tracking and sensor fusion (Tracking and Sensor Fusion)

Multi-sensor and object tracking algorithm to provide on the basis of bird's-eye view of the object detection and tracking.

2.5 terrain contour matching (Geographic Mapping)

2.5.1 get online map HERE (Accessing HERE HD Live Map Data)

Read HERE network services through direct ( foreign map providers ) high-definition map, the map information includes detailed road, lane and positioning information.

2.5.2 Map visualization

Use streaming coordinate drawn position when the vehicle is traveling.

2.6 planning and control decision-making

Use of vehicles costmap and motion planning algorithms driving route planning.

summary

Wanted to make a profile in accordance with the relevant articles and personal circumstances of the trial MATLAB toolbox official gateway to the autopilot, but read, read, write discovered that in fact the official website has been very full, and just two articles generals net mixed together in half translation down, try to talk to the follow-on article.

reference

The main reference to the official website two documents:

https://www.mathworks.com/products/automated-driving.html#sft

https://www.mathworks.com/help/driving/index.html?s_tid=CRUX_lftnav