Tianracer

中文版说明
TianRacer详细中文操作手册
Tianracer is a series of low-cost autonomous vehicles ranging in size from 1:10 to 1:5, controlled in a closed-loop system, allowing them to follow input commands more faithfully. All Tianracer models are equipped with NVIDIA development boards, including Jetson Nano, Jetson TX2, Xavier, Orin, and more. The basic version of Tianracer can reach a maximum speed of 3m/s, while the pro version can achieve speeds up to 10m/s.

Introduction

TianRacer ROS Wiki Our objective is to develop an affordable autonomous racing car, the Tianracer, with capabilities extending beyond basic mapping, localization, and navigation. It will feature advanced functionalities such as object detection, traffic light detection, and lane keeping. Currently, Tianracer builds upon and enhances the AI capabilities inherited from Jetracer.

Purchase from Tianbot Official Taobao Store

The Tianracer Nano version is a standard platform with all the hardware and software installed, a ready-to-go racing car.

点击这里进入淘宝购买或咨询客服： Purchase from Taobao:

Appreciation to HyphaROS RaceCar and Lord-Z

We developed the Tianracer based on the Hypha racecar. Please check out the original authors' repo for more details. https://github.com/Hypha-ROS/hypharos_racecar

Developer:

• HaoChih, LIN
• KaiChun, WU

Tianracer gazebo simulation is imported from https://github.com/Lord-Z/ackermann_gazebo

Developer:

• Yuxing, ZHANG

Customization

Speed: 3m/s - 10m/s Control：closed-loop speed control Computer: Nvidia Jetson Nano/TX/NX/Orin Developer Kit Chassis: BLDC Motor + TianBoard Mini + Servo Lidar: Slamtec / Osight / Richbeam / Livox Camera: 1080P Fisheye Undistorted USB3.0 / RGBD Camera Remote Controller: DJI DT7 Battery: LiPo

Instructions

Installation

cd ~/catkin_ws/src/
git clone https://github.com/tianbot/tianracer.git
cd ~/catkin_ws && catkin_make

Simulation

The main simulation workflow in this repository is based on Gazebo.

If you want to bring up the full simulation stack first, start with:

roslaunch tianracer_gazebo demo_tianracer_teb_nav.launch

In this repository, launch files starting with demo_ are usually the easiest end-to-end entry points. This one brings up:

• Gazebo world
• map server
• robot spawn
• controllers
• TEB navigation
• AMCL localization
• RViz

If you only want to check whether the simulated car, controllers, odometry, and TF are working normally, run:

roslaunch tianracer_gazebo tianracer_bringup.launch

This launch mainly starts:

• tianracer_on_racetrack.launch
• tianracer_control.launch

It is a better starting point when you only want to confirm that:

• the Gazebo world starts correctly
• the robot is spawned correctly
• the controllers are loaded
• odometry and TF are being published normally

For multi-robot simulation, start with:

roslaunch tianracer_gazebo swarm/demo_sim_two_tianracer.launch

If you only want two cars in Gazebo without navigation, run:

roslaunch tianracer_gazebo swarm/two_tianracer_bringup.launch

The related packages are organized as follows:

• tianracer_gazebo: simulation worlds, robot model, controllers, and Gazebo demo launches
• tianracer_sim_referee: simulation-only race timing and judging runtime
• tianracer_navigation: navigation algorithms and generic navigation launches
• tianracer_rviz: generic RViz configurations
• tianracer_description: meshes and TF-related model resources

For now, the recommended simulation entries in this repository are:

• tianracer_gazebo/launch/demo_tianracer_teb_nav.launch
• tianracer_gazebo/launch/tianracer_bringup.launch
• the tianracer_gazebo/launch/swarm/ launches

The following files should not be treated as the main simulation entry points for now:

• tianracer_gazebo/launch/navigation.launch
• tianracer_gazebo/launch/tianracer_gmapping.launch
• tianracer_gazebo/launch/spawn_model.launch

Simulation referee

The referee is an independent ROS package rather than a Gazebo plugin. Start the simulation first, then launch the referee in another terminal:

roslaunch tianracer_sim_referee referee.launch

It observes Gazebo's /clock and /gazebo/model_states topics and does not send navigation goals or vehicle-control commands. The package exposes a small public launcher and ships the referee business logic as _referee_core.so only. Official gate geometry is versioned and compiled into the core. The GUI automatically identifies the actual course and vehicle from Gazebo, independently of shell world/robot variables. Only verified tianracer_racetrack runs are submittable; other official courses remain local practice. Unsupported worlds are rejected rather than loading a customer-editable checkpoint file.

Interfacing

Tianracer can be brought up all at once, or separately.

roslaunch tianracer_bringup tianracer_bringup.launch

Tianracer Base

roslaunch tianracer_core tianracer_core.launch

Lidar

roslaunch tianracer_bringup lidar.launch

RGBD Camera (if applicable)

roslaunch tianracer_bringup rgbd_camera.launch

USB Camera

roslaunch tianracer_bringup usb_cam.launch

GPS (if applicable)

roslaunch tianracer_bringup gps.launch

Mapping

If you want to build a map in the current simulation workflow, bring up Gazebo first and confirm that the car, lidar, odometry, and TF are all normal:

roslaunch tianracer_gazebo tianracer_bringup.launch

After Gazebo is running, choose one SLAM method below.

GMapping

This is the most direct 2D mapping entry in the current repository:

roslaunch tianracer_slam tianracer_gmapping.launch

HectorSLAM

If you want to compare another 2D mapping pipeline, run:

roslaunch tianracer_slam tianracer_hector.launch

Cartographer

For Gazebo, use the Gazebo-specific Cartographer launch so that use_sim_time is configured for simulation:

roslaunch tianracer_slam gazebo/gazebo_cartographer_2d.launch

The generic launch below is still available, but it is more suitable when you already know the time source and topic setup you want to use:

roslaunch tianracer_slam tianracer_cartographer.launch

Save the Map

If the map is generated in Gazebo, it is better to save it with the Gazebo-specific launch. By default, the map will be written to tianracer_gazebo/maps/ and the file prefix is racetrack:

roslaunch tianracer_slam gazebo/gazebo_map_save.launch

The generic save entry is still available and writes to tianracer_slam/maps/ with the default prefix tianbot_office:

roslaunch tianracer_slam map_save.launch

Navigation

If you only want to run the complete Gazebo navigation demo, the simplest entry is still:

roslaunch tianracer_gazebo demo_tianracer_teb_nav.launch

That launch already brings up the Gazebo world, map server, localization, navigation, and RViz together.

If you already have a map and want to run the navigation stack separately, then use the launches in tianracer_navigation.

Teb Local Planner

This is the main map-based navigation entry in tianracer_navigation:

roslaunch tianracer_navigation tianracer_teb_nav.launch use_rviz:=true map_file:=tianbot_office

This launch mainly starts:

• map_server
• AMCL
• move_base
• cmd_vel_to_ackermann_drive.py
• RViz (optional)

If you are using Gazebo, bring up the simulated car first:

roslaunch tianracer_gazebo tianracer_bringup.launch

L1_Controller Local Planner

If you want to test the L1 local planner on top of the same map/localization flow, run:

roslaunch tianracer_navigation tianracer_l1_nav.launch map_file:=tianbot_office use_rviz:=true

This launch keeps the same map server and AMCL path, but switches the local planning side to the L1 controller flow.

If you only want to open the generic RViz view from another machine after ROS networking is configured, run:

roslaunch tianracer_rviz view_teb_planner.launch

Reactive Navigation

The repository also keeps two reactive navigation entries that are useful for quick experiments in Gazebo.

If you are testing in Gazebo, bring up the simulated car first:

roslaunch tianracer_gazebo tianracer_bringup.launch

wall_following

Then run:

roslaunch tianracer_navigation wall_following.launch use_rviz:=true

This launch mainly starts:

• wall_following.py
• ackermann_convert_drive.py
• RViz (optional)

This wall_following entry is now organized as a reactive controller demo: it only depends on lidar input, publishes ackermann_cmd_stamped, and uses ackermann_convert_drive.py to keep the existing ackermann_cmd output chain compatible. It does not bring up a map or AMCL.

Treat this launch as a lightweight reactive demo entry rather than a full navigation stack replacement.

follow_the_gap

Then run:

roslaunch tianracer_navigation follow_the_gap.launch use_rviz:=true

This launch mainly starts:

• tianracer_gazebo/follow_the_gap.py
• ackermann_convert_drive.py
• RViz (optional)

This one has the same caveat: although it is grouped here as no-map navigation, the current launch file still includes map_server and AMCL. For now, it should also be understood as a reactive navigation demo built on top of the current debugging layout.

License: GPL v3