Skip to content
Go to file

Latest commit


Git stats


Failed to load latest commit information.
Latest commit message
Commit time

End-to-end Autonomous Driving Perception

[Project webpage] [Paper]

This repo contains code for End-to-end Autonomous Driving Perception with Sequential Latent Representation Learning. This work introduces a novel end-to-end approach for autonomous driving perception. A latent space is introduced to capture all relevant features useful for perception, which is learned through sequential latent representation learning. The learned end-to-end perception model is able to solve the detection, tracking, localization and mapping problems altogether with only minimum human engineering efforts and without storing any maps online. The proposed method is evaluated in a realistic urban driving simulator (CARLA simulator), with both camera image and lidar point cloud as sensor inputs.

System Requirements

  • Ubuntu 16.04
  • NVIDIA GPU with CUDA 10. See GPU guide for TensorFlow.


  1. Setup conda environment
$ conda create -n env_name python=3.6
$ conda activate env_name
  1. Install the interpretable end-to-end driving package following the installation steps 2-4 in

  2. Clone this git repo to an appropriate folder

$ git clone
  1. Enter the root folder of this repo and install the packages:
$ pip install -r requirements.txt
$ pip install -e .


  1. Enter the CARLA simulator folder and launch the CARLA server by:
$ ./ -windowed -carla-port=2000

You can use Alt+F1 to get back your mouse control. Or you can run in non-display mode by:

$ DISPLAY= ./ -opengl -carla-port=2000

It might take several seconds to finish launching the simulator.

  1. Enter the root folder of this repo and run:
$ ./

It will then connect to the CARLA simulator, collect driving data, then train and evaluate the agent. Main Parameters are stored in params.gin.

  1. Run tensorboard --logdir logs and open http://localhost:6006 to view training and evaluation information.

Trouble Shootings

  1. If out of system memory, change the parameter replay_buffer_capacity and initial_collect_steps of the function tran_eval smaller.

  2. If out of CUDA memory, set parameter model_batch_size or sequence_length of the function tran_eval smaller.


If you find this useful for your research, please use the following.

  title={End-to-end Autonomous Driving Perception with Sequential Latent Representation Learning},
  author={Chen, Jianyu and Xu, Zhuo and Tomizuka, Masayoshi},
  journal={arXiv preprint arXiv:2003.12464},


End-to-end Autonomous Driving Perception with Sequential Latent Representation Learning




No releases published


No packages published
You can’t perform that action at this time.