Source: Transit State
Write on the front:
Autonomous vehicles need to undergo extensive road testing to meet commercial requirements before they can be truly commercially adopted. The time and cost of using road testing to optimize autonomous driving algorithms are too high, and open road testing is still restricted by regulations, extreme traffic conditions and scene reproduction are difficult, and there are hidden dangers in test safety. At present, autonomous driving simulation testing has been widely accepted by the industry, and about 90% of the autonomous driving algorithm test is completed through the simulation platform, 9% is completed in the test field, and 1% is completed through the actual road test.
The autonomous driving simulation test platform must have several core capabilities: real-life restoration of test scenarios, efficient use of road acquisition data to generate simulation scenarios, massive parallel acceleration in the cloud, etc., so that simulation tests can meet the closed-loop of autonomous driving perception, decision planning and control full-stack algorithms. At present, including technology companies, car companies, autonomous driving solution providers, simulation software companies, universities and scientific research institutions and other entities are actively involved in the construction of virtual simulation platforms.
This article details the existing autonomous driving simulation software for the reader's reference, and the software ranks in no particular order.
0、CarSim
CarSim, as well as the related TruckSim and BikeSim are powerful dynamic simulation software developed by Mechanical Simulation and are widely used by OEMs and suppliers around the world. CarSim is aimed at four-wheeled cars, light trucks, TruckSim is aimed at trucks with multi-axles and twin tires, and BikeSim is aimed at two-wheeled motorcycles. CarSim is a vehicle dynamics simulation software, mainly from the vehicle perspective of simulation, it built a considerable number of vehicle mathematical models, and these models have a wealth of experience parameters, users can quickly use, eliminating the complicated process of modeling and parameter adjustment.
The CarSim model can run 10 times faster than real-time on the computer, can simulate the response of the vehicle to the driver control, 3D road surface and aerodynamic input, the simulation results are highly close to the real vehicle, mainly used to predict and simulate the handling stability, braking, smoothness, power and economy of the whole vehicle. CarSim comes with a standard Matlab/Simulink interface, which can be easily combined with Matlab/Simulink for controlling the development of algorithms, and can generate a large number of data results for subsequent analysis or visualization using Matlab or Excel during simulation. CarSim is also available in an RT version that can support mainstream HIL test systems such as dSpace and NI systems for easy joint HIL simulation.
CarSim is also supported by ADAS-related functions, which can build parametric road models, more than 200 moving traffic objects, using scripts or controlling their movements externally via Simulink, while adding up to 99 sensors to detect moving and stationary objects. The recent version of CarSim has been enhanced in the development of ADAS and autonomous driving, adding more 3D resources such as traffic signs, pedestrians, etc., as well as the import process of high-definition maps. CarSim also provides a Unreal engine plug-in that can be simulated in conjunction with the Unreal engine.
1、CarMaker
Carmaker, along with the associated TruckMaker and MotoriccleMaker, is a simulation software for dynamics, ADAS and autonomous driving launched by the German company IPG. Carmaker is first and foremost an excellent dynamic simulation software that provides accurate vehicle body models (engine, chassis, suspension, transmission, steering, etc.), in addition to creating a closed-loop simulation system that includes vehicles, drivers, roads, and traffic environments.
IPGRoad: Can simulate multiple lanes, intersections and other forms of roads, and can be configured GUI to generate conical, cylindrical and other forms of barricades. The geometry of the road and the condition of the pavement (unevenness, roughness) can be arbitrarily defined.
IPGTraffic: Is a traffic environment simulation tool, providing a rich range of traffic objects (vehicles, pedestrians, road signs, traffic lights, road construction buildings, etc.) model. Simulation of real traffic environments is possible. The test vehicle can identify the traffic object and trigger the action (e.g. the speed limit sign can trigger the vehicle to slow down accordingly).
IPGDriver: Advanced, self-learning driver model. Vehicles can be controlled under a variety of driving conditions, such as starting uphill, parking in storage, and flicking the steering wheel. And can adapt to the vehicle's dynamic characteristics (drive form, transmission type, etc.), road friction coefficient, wind speed, traffic environment conditions, adjust the driving strategy.
As a platform software, CarMaker can integrate with many third-party software, such as ADAMS, AVLCruise, rFpro, etc., and can take advantage of the advantages of each software for joint simulation. At the same time, CarMaker's hardware provides a large number of board interfaces, which can be easily tested with ECU or sensors.
2、PreScan
PreScan is an ADAS test simulation software developed by TassInternational and acquired by Siemens in August 2017. PreScan is an emulation platform consisting of a GUI-based preprocessor for defining the scene and a running environment for executing the scene. The main interfaces engineers use to create and test algorithms include MATLAB and Simulink. PreScan can be used for applications ranging from model-based controller design (MIL) to real-time testing using software-in-the-loop (SIL) and hardware-in-the-loop (HIL) systems.
The PreScan can operate in open-loop, closed-loop, offline and online modes. It is an open software platform with a flexible interface that connects to third-party vehicle dynamics models (e.g., CarSIM and dSPACEASM) and third-party HIL simulators/hardware (e.g., ETAS, dSPACE, and Vector).
Prescan consists of multiple modules, which are mainly divided into four steps: building a scene, adding sensors, adding control systems, and running simulations.
Scene Construction: PreScan provides a powerful graphical editor that allows users to build rich simulation scenarios using road segmentation, including traffic signage, a library of basic components for trees and buildings, including a library of traffic participants for motor vehicles, bicycles, and pedestrians, modifying weather conditions (such as rain, snow, and fog), and light sources (such as sunlight, headlights, and street lights). The new version of PreScan also supports the import of high-definition maps in OpenDrive format for creating more realistic scenes.
Add Sensors: PreScan supports a wide range of sensors including Ideal Sensors, V2X Sensors, LiDAR, Millimeter Wave Radar, Ultrasonic Radar, Monocular and Binocular Cameras, Fisheye Cameras, etc. Users can add as much as they want.
Add control systems: Control models can be established through MATLAB/Simulink, or closed-loop control can be performed with third-party dynamic simulation models (e.g., CarSim, VI-Grade, dSpace ASM's vehicle dynamics models).
Run experiments: The 3D visual viewer allows users to analyze the results of their experiments while providing image and animation generation capabilities. In addition, the interface using ControlDesk and LabView can be used to automatically run experimental batch scenarios as well as run hardware-in-the-loop simulations.
3、PTV Vissim
Vissim is one of the world's leading microscopic traffic flow simulation software provided by the German company PTV. Vissim can easily construct a variety of complex traffic environments, including highways, large roundabouts, parking lots, etc., and can also simulate the interaction behavior of motor vehicles, trucks, rail traffic and pedestrians in a simulation scenario. It is an effective tool for professional planning and evaluation of urban and suburban transportation facilities, and can also be used to simulate the impact of local emergency traffic, the evacuation of a large number of pedestrians, etc.
Vissim's simulations can achieve high precision, including microscopic individual stalking and lane-changing behavior, as well as group cooperation and conflict. Vissim has built-in analytical tools to obtain a variety of specific data results in different situations or to obtain intuitive understanding from a high-quality 3D visualization engine. Driverless algorithms can also be simulated and tested using simulated highly dynamic traffic environments by plugging into Vissim.
4、TESS
TESS simulation system is the first generation of road traffic simulation system developed by Professor Sun Jian of Tongji University in 2006. Since then, after ten years, Professor Sun Jian's research group has carried out more than 100 model innovations and simulation system application practices for the operation characteristics of mixed traffic flow in China. The main functions of TESS NG micro traffic simulation system are: full traffic scene simulation, multi-mode traffic simulation, intelligent transportation system simulation, visual evaluation, secondary development interface, support for 3D scene display, etc. At the same time, TESS NG can be integrated with urban traffic brains, traffic control systems, computable road networks (such as OpenDrive, OpenStreetMap, etc.), and can be integrated with driving simulators, BIM/CIM systems, intelligent car virtual test tools, etc. to achieve cross-industry applications. Users can also achieve more cross-industry applications through customized services.
5、SUMO
SUMO is an open source microscopic continuous traffic flow simulation software developed by the German National Aerospace Center. It comes with a traffic simulation road network editor that can be interactively edited to add roads, edit lane connections, handle intersection areas, edit signal timing, etc. It is also possible to convert road networks from Vissim, OpenStreetMap, openDrive via a separate conversion program. You can specify the route for each vehicle by editing the route file, or you can use parameters to generate randomly. At runtime, it can simultaneously handle the continuous traffic simulation requirements of several square kilometers and tens of thousands of vehicles, and also provides an OpenGL-based visualization to display the results of traffic simulation in real time.
In addition, SUMO provides convenient C++ and Matlab interfaces that can be flexibly run in conjunction with third-party emulators. SUMO itself is used as a traffic field traffic, timing, prediction and other simulations, and has recently begun to be applied to the simulation of unmanned driving, providing a random complex dynamic environment for driverless algorithms.
6、VIRES VTD
VTD (VirtualTest Drive) is a complete modular simulation toolchain for ADAS, active safety and autonomous driving developed by VIRES in Germany. VIRES was acquired by MSC Software Group in 2017. VTD currently runs on the Linux platform, and its functions cover road environment modeling, traffic scene modeling, weather and environment simulation, simple and physically realistic sensor simulation, scene simulation management, and high-precision real-time screen rendering. Full-cycle development processes from SIL to HIL and VIL can be supported, and the open modular framework can be easily simulated in conjunction with third-party tools and plug-ins. VIRES is also a major contributor to the widely used open formats for autonomous driving simulation, OpenDrive, OpenCRG, and OpenScenario, on which VTD's capabilities and storage also rely. The simulation process of VTD is mainly composed of three steps: road network construction, dynamic scene configuration, and simulation operation.
1) VTD provides a graphical interactive road network editor, the RoadNetwork Editor (ROD), which can simultaneously generate OpenDrive high-definition maps while building complex road simulation environments with multiple types of lanes using various traffic elements.
2) In the establishment of dynamic scenes, VTD provides a graphical interactive scene editor, ScorelioEditor, which provides traffic bodies that add user-defined behavior control based on OpenDrive, or traffic flows running continuously in an area.
3) Whether it is SIL, or HIL, whether it is real-time or non-real-time simulation, whether it is a stand-alone or high-performance computing environment, VTD provides the corresponding solution. VTD runtime can simulate real-time high-quality light and shadow effects and road reflections, body rendering, rain, snow and fog weather rendering, sensor imaging rendering, big light visual effects, etc.
7、rFpro
rFpro is a British company founded in 2008 as an in-house track reconstruction and simulation project for F1 teams, which also determined that it had high requirements for the speed, real-time and accuracy of the simulation from the beginning. rFPro uses high-precision phased lidar to scan data on the road surface and shoulder, which can generate a high-precision digital model of the road surface with a resolution of 1cm, while using TOF lidar to scan the streets and scenes on the roadside, in this way it can provide dynamic simulation, ADAS, and autonomous driving tests with virtual scenes that are highly matched to the real environment. rFpro uses this method to create high-precision virtual scenes for numerous tracks and test scenarios, including F1, NASCAR, IndyCar, and more.
In terms of dynamic scene simulation, rFpro can plug into SUMO or Vissim, which generates a continuous flow of traffic to fill the entire scene, or it can co-simulate with Carmaker to provide more realistic sensors and pavement inputs for Carmaker's test scenes. rFpro also provides a physically realistic lighting and weather system that can effectively simulate changes in sky light and rain, fog and other weather.
8、Sister-in-law
Cognata, an Israeli self-driving simulation startup founded in 2016, completed a $18.5 million Series B funding round at the end of 2018. Using a combination of artificial intelligence, deep learning and computer vision, Cognata recreates the city on its 3D simulation platform, providing customers with a variety of test scenarios that simulate real-world test driving.
Cognata's technology is mainly divided into three aspects, in terms of static environments, Cognata's TrueLife3DMesh engine uses computer vision and deep learning algorithms to automatically generate a virtual simulation environment including buildings, roads, lane signs, traffic signs based on maps and satellite images. In terms of dynamic simulation, Cognata builds accurate and scalable traffic simulation models and weather lighting models based on street historical traffic data, simulating a variety of vehicles and pedestrians in real-world environments. The entire virtual simulation engine combines static and dynamic simulation models to simulate the interaction of sensors with various changes in the simulated environment, providing a complete feedback loop for the autonomous driving system to be tested.
Cognata's simulation technology is powered by NVIDIA's DGX Station, and in March 2019, Cognata announced a partnership with NVIDIA to leverage its powerful computing power on NVIDIA's platform to simulate multiple virtual vehicles in a virtual environment for large-scale testing.
9、RightHook
RightHook is a California-based startup that provides simulation solutions for the autonomous driving industry. RightHook offers a full set of toolchains, including RightWorld, RightWorldHD, RightWorldHIL, and more. RightWorld provides a process for automatically reconstructing a virtual scene with rich details from a high-definition map, and provides a simple and easy-to-use test case creation process, which can be organically extended by AI algorithm after the case is created. RightWorld also offers a deterministic intelligent traffic simulation model that includes vehicles, pedestrians, and bicycles.
RightWorldHD simulates dynamics, weather, time changes and sensors (including cameras, Lidar, Radar, IMU and GPS) while supporting a rich set of interfaces including NVIDIADriveWorks, LCM and ROS. RightWorldHIL provides support for HIL testing with a mix of software, algorithms, and hardware.
10、ParallelDomain
ParallelDomain is a startup founded in California, USA in 2017. At the end of 2018, ParallelDomain received investment from Toyota. ParallelDomain strives to automatically generate high-quality virtual environments, and its software can automatically generate the city blocks for the required tests in a very short time.
The ParallelDomain platform uses real-world map data, can receive a variety of map formats, uses additional elements where the map does not provide sufficient data, and relies on a programmatic generation engine to automatically generate virtual worlds. A notable feature is that all elements of the virtual world are adjustable and programmable, such as the number of lanes, terrain type, mountain range location, road curvature, etc. ParallelDomain also provides dynamic traffic scenes for auto-generated scenes.
11、51Sim-One
51Sim-One is an autonomous driving simulation and test platform independently developed by 51VR that integrates multi-sensor simulation, traffic flow and agent simulation, perception and decision simulation, and autonomous driving behavior training. The simulation platform is based on the mechanism modeling of physical characteristics, with the characteristics of high precision and real-time simulation, which is used for the research and development, testing and verification of autonomous driving products, which can quickly accumulate automatic driving experience for users, ensure product performance safety and reliability, improve product development speed and reduce development costs.
In terms of scene construction, You can quickly create an OpenDrive-based road network from scratch through WorldEditor, or restore road network information from real data such as point cloud data and map imagery. Supports importing existing OpenDrive format files for secondary editing, and finally 51Sim-One automatically generates the required static scenes. Support the free configuration of global traffic flow, independent traffic agents, opponent vehicles, pedestrians and other elements in the scene to build dynamic scenes, combined with the simulation of lighting, weather and other environments to present a rich and changeable virtual world.
In terms of sensor simulation, the 51Sim-One supports multi-channel simulation of common type or custom requirements sensors to meet the testing and training of perceptual system algorithms, and also supports various hardware-in-the-loop test requirements. For camera simulation, 51Sim-One provides annotated image datasets such as semantic segmentation maps, depth maps, 2D/3D enveloping boxes, monocular, wide angle, fisheye and other camera simulations. For radar simulation, the raw data of the lidar point cloud, the data of the marked point cloud, the bounding box of the identified object can also provide the target-level millimeter wave radar detector data.
12・Pilot-DGaiA
GaiA is a verification simulation tool developed by Peidai (Shanghai) for autonomous driving and ADAS. It can restore complex roads by integrating the road network database, and reproduce the realistic driving environment through the use of the environmental building model library. GaiA offers a wide range of C++ and matlab interfaces for a wide range of vehicles and systems under test. GaiA can generate a large number of traffic participants and set them manually and automatically for their traffic behavior planning, and can even change the aggressiveness of driving behavior. GaiA also offers high-fidelity environmental perception sensors including millimeter-wave radar, lidar, cameras, etc.
13、Metamoto
Founded in 2016, Metamoto is a Silicon Valley startup. Metamoto provides "Simulationas a Service" for self-driving companies, trying to help self-driving companies achieve iterations of development through an accelerated feedback loop. Its products mainly include three parts: designer, cloud platform and analyzer. The designer can be used to add road networks, other environmental vehicles, pedestrians and traffic lights to build a test scenario that can generate multiple test cases by controlling the range of values for various parameters.
The cloud platform is responsible for scheduling hardware resources according to the situation of the test case, running the test case in parallel, and generating a large amount of test data. After the operation is completed, the simulated sensor data and various simulation information of the vehicle can be used to debug the autonomous driving system using the analyzer. Metamoto supports accurate simulation of various sensors including lidar, camera, millimeter wave radar, ultrasonic radar, GPS, IMU, etc., and can react differently to different materials. A significant feature of Metamoto is that it provides a quick way to adjust and override the parameters of the test, and can run a large number of tests in a short period of time with the support of the cloud platform, which effectively improves the test efficiency.
14、ESIPro-Sivic
ESI's Sensor Simulation Analysis Solution Pro-SiVIC helps manufacturers in the transportation industry to virtually test the performance of multiple sensing systems on board or on board, and to accurately reproduce influencing factors such as lighting conditions, weather, and other road users.
Pro-SiVIC can be used to create highly realistic, 3D scenes equivalent to real-world scenes, and realize real-time interactions in the scene for simulation analysis, reducing the need for physical prototypes. Customers can quickly and accurately simulate the performance of individual embedded systems in typical and extreme operating environments, and it can provide sensor models based on multiple technologies, such as: cameras, radar, lidar (laser scanners), ultrasonic sensors, GPS, odometers and communication equipment. Taking the automotive industry as an example, Pro-SiVIC offers several environmental catalogues that provide representative representation of different roads (urban roads, highways and country roads), traffic signs and lane markings.
15、NVIDIADrive Constellation
NVIDIADrive Constellation is NVIDIA's autonomous driving simulation platform, which is mainly composed of two parts in hardware, one is a DGX server, which runs the DriveSim software system, relying on the powerful graphics computing power of DGX, truly simulating the lighting, night and various weather changes in the actual environment, and the other server is equipped with the DRIVEAGX Pegasus vehicle computer. The algorithm used to run the full stack of autonomous driving, the two parts form a complete HIL simulation closed loop.
16、PanoSim
PanoSim is a simulation software platform that integrates complex vehicle dynamics model, car three-dimensional driving environment model, automobile driving traffic model, vehicle environment sensing model (camera and radar), wireless communication model, GPS and digital map model, Matlab/Simulink simulation environment automatic generation, graphics and animation post-processing tools. It is based on the principle of physical modeling and numerical simulation that takes into account both precision and efficiency, realistically simulates various environments and working conditions of automobile driving, and establishes a high-precision camera, radar and wireless communication model based on the concept of combining geometric model and physical modeling to support the research and development, testing and verification of technologies and products such as automobile dynamics and performance, automotive electronic control system, intelligent assisted driving and active safety system, environmental sensing and perception, and automatic driving in the digital simulation environment.
PanoSim includes not only complex vehicle dynamics models, chassis (braking, steering and suspension), tires, drivers, powertrains (engines and transmissions), but also models and simulation analyses of large, medium and small cars in a variety of typical drive types and suspension forms. It provides 3D digital virtual test scene modeling and editing capabilities, supporting the modeling and editing of roads and road textures, lane lines, traffic signs and facilities, weather, night scenes and other car driving environments.
17. AAI
AAI (Automotive Artificial Intelligence) is a startup founded in Berlin in 2017. AAI has built a complex set of high-simulation virtual environments created based on high-precision maps, which will use artificial intelligence technology to integrate traffic participants into the virtual simulation environment, and use real-life driving behavior data to train participant behavior using machine learning algorithms to generate driver profiles such as aggressive drivers, gentle drivers, and defensive drivers, with the goal of replicating the real world and realistically simulating all road users and environmental factors. AAI supports a variety of sensor simulations and also provides an analyzer for in-depth analysis of the data generated by the simulation.
18、AirSim
AirSim is an open source Drone and Autonomous Driving Simulation Research Project built on UnrealEngine by Microsoft Research. AirSim is implemented as an Unreal Engine plug-in, which makes full use of Unreal Engine's ability to create a highly realistic virtual environment, can simulate shadows, reflections and other real-world environments, and the virtual environment can easily generate a large number of annotation data capabilities, while providing a simple and convenient interface, can allow drones and autonomous driving algorithms to access a lot of training. The main goal of AirSim is to serve as a platform for AI research to test end-to-end reinforcement learning algorithms for deep learning, computer vision, and autonomous vehicles. The latest AirSim also offers a version of the Unity engine, adding lidar support.
19、CARLA
CARLA is an open source simulator developed under the guidance of the Computer Vision Center of the Autonomous University of Barcelona, Spain, for the development, training and validation of autonomous driving systems. Like AirSim, Carla was developed using Unreal Engine, using a server and multiclient architecture. In terms of scenarios, CARLA provides open source digital resources (including urban layouts, buildings, and vehicles) that create scenarios for autonomous driving, as well as several scenarios built from these resources for autonomous driving test training. At the same time, CARLA can also use VectorZero's road building software RoadRunner to create scenes and supporting high-definition maps, and also provides a simple map editor. CARLA can also support the flexible configuration of sensors and environments, it supports multi-camera, lidar, GPS and other sensors, and can also adjust the lighting and weather of the environment. CARLA provides simple automatic behavior simulation of vehicles and pedestrians, and also provides a complete set of Python interfaces that can control vehicles, signal lights, etc. in the scene, which is used to facilitate joint simulation with automatic driving systems, complete decision-making systems and end-to-end reinforcement learning training.
20、LGSVL Simulator
LGSVLSimulator is an open source autonomous driving simulator developed by LG's Silicon Valley Labs based on the Unity engine. It provides integration with open source autonomous driving platforms Autoware and BaiduApollo. Users can annotate based on 3D scenes within Unity and export them to a high-definition map format that matches the automated driving system. It also provides support for sensor simulation including lidar, millimeter wave radar, GPS, IMU, cameras, and can simultaneously output the original results and truth values of the sensors.
21, Baidu Apollo
As an important part of Baidu Apollo platform, Baidu Apollo simulation platform is used to support the development and iteration of internal Apollo systems on the one hand, and provide cloud-based decision system simulation services for developers of the Apollo ecosystem on the other hand. The Apollo simulation platform is a cloud service built on Baidu Cloud and Azure that can be tested in the cloud using user-specified versions of Apollo. The Apollo simulation scenario can be divided into Worldsim and Logsim. Worldsim is a scene composed of artificially preset roads and obstacles, which can be used as a unit test to test autonomous vehicles simply and efficiently, while Logsim is a scene extracted from road test data, which truly reflects the complex and changeable obstacles and traffic conditions in the actual traffic environment. The Apollo simulation platform also provides a relatively complete scene through the discriminant system, which can evaluate the automatic driving algorithm from the aspects of traffic rules, dynamic behavior and comfort.
Apollo has also partnered with Unity to develop Unity-based realistic virtual environment simulations that can provide 3D virtual environments, roads, and weather changes. Recently, Baidu has also proposed a new data-driven approach for end-to-end simulation of autonomous driving: Enhanced Autonomous Driving Simulation (AADS). This method leverages simulated traffic flows to enhance real-world images to create realistic simulated scenes that resemble photos rendered in the real world. Specifically, it is recommended to scan Street View with LiDAR and camera. Decompose the input data into background, scene lighting, and foreground objects. At the same time, a new view compositing technique is proposed, which can change the viewpoint on a static background. The foreground vehicle is equipped with a computer 3D model. With precisely estimated outdoor lighting, 3D vehicle models, computer-generated pedestrians, and other movable subjects can be repositioned and rendered back to the background image to create a realistic Street View image. In addition, simulating traffic flow, placement and movement of synthetic objects, capturing real-world vehicle trajectories that look natural and capture the complexity and diversity of real-world scenes.
22、Waymo Carcraft
Representing the world's leading level of Waymo unmanned vehicles, a core secret is its Carcraft simulator, which is the key to Waymo's unmanned vehicles being able to travel billions of miles a year. At the beginning of Carcraft's development, the system was only used to visually replay the situation of roadside vehicles on the road, and since then it has played an increasingly important role. Carcraft can test each new software version using real-world driving playback data to verify algorithm improvements, discover new problems, and build entirely new virtual scenarios for testing. Every day, 25,000 virtual Waymo driverless vehicles travel more than eight million miles in the simulator to reinforce existing self-driving skills and test new ones. The biggest advantage of simulation testing is that it can quickly repeat some scenarios that do not often occur in reality but are very important, such as five forks in the road and merger into the roundabout. The simulator gives the self-driving system many opportunities to practice this single scenario to master the corresponding skills. In addition, in the simulator, you can make some changes to a participant in a specific test scene, or a traffic signal, add additional pedestrians, etc., in this way you can build a large number of derivative scenes, so as to test the driverless algorithm more fully.
23. Tencent TAD Sim simulation platform
At the beginning of the design, Tencent's automatic driving virtual simulation platform TAD Sim is different from the traditional simulation system, which is specially designed and developed for automatic driving test verification, built-in centimeter-level high-precision map, built a digital twin system containing dynamic and static elements, and tested the completeness of the automatic driving algorithm with ever-changing scenarios.
Part of the content comes from Zhou Xiqin and the Blue Book of China's Autonomous Driving Simulation Technology 2020, special thanks