To build a self driving car simulator for Chinese traffic, which based on intel open source simulator Carla.

There are several targets show below:

• Integrate perception, localization, path planning and control in Carla simulator.
• Setup Chinese city traffic map
• Verify multi-camera + radar solution feasibility
• Setup metrics to evaluate self driving car algorithm

Reproduce Carla modular pipeline

1. Carla installation and learning

• Receive measurement data from server
• Send control data to control vehicle
• Define radar sensor
• Reproduce Carla team modular pipeline.

2. Integrate perception, localization, path planning and control

• Lane detection
• Traffic Signs detection
• Vehicles, pedestrian detection
• Localization
• Path planning
• Vehicle control
• Integrating

3. Customize Chinese city map
4. Verify multi-camera + radar solution feasibility
5. Evaluate self driving car algorithm

• Ubuntu 16.04
• ROS
• Python 3.5
• Anaconda
• Tensorflow 1.4.0

• Carla github
• Carla document
• Carla 0.7 baidu pan 链接: https://pan.baidu.com/s/1eSuBh5K 密码: dgqz
• Carla introduction zhihu
• Carla paper
• Carla tutorial
• Udacity self driving car simulator
• Airsim

• Traffic lights filter if there are several detections.

Examples: if model detects two green lights from front camera images, there should be a pipeline to determine which light should be used.

• Vehicle dynamics simulation improvements
• Vehicle customization

The core contributors are a team including several Chinese Udacity self driving car Nanodegree graduates.

CARLA Self Driving Car Simulator in Chinese Traffic Scenes specific code is distributed under MIT License.

Related assets follows CARLA Licenses