Road Detection with Point Cloud Library and ZED Stereoscopic Vision System

Introduction Road detection plays an integral role in self-driving cars. Accuracy and reliable road detection can pave the road for good path planning. In the self-driving golf cart project, I use two methods to perform road detection. 1. I use semantic segmentation and deep learning to classify each pixel in an image. 2. I use… Continue reading Road Detection with Point Cloud Library and ZED Stereoscopic Vision System

Self-Driving Software + Carla Simulator

Want to learn more about self-driving car simulators? Check out this blog post. Simulators are crucial in developing a good self-driving system. Not everyone will have access to quality self-driving hardware platforms, therefore software simulator is a nature incubator for cutting-edge research and development.

GPS Localization with ROS, rviz and OSM

When you place a robot in a known environment, the robot should be able to localize itself, meaning it could compute the approximate position of itself using sensors and algorithms. Localization is crucial to self-driving vehicles, which requires incredible precision. GPS (Global Positioning System) is one of the first steps to localization. However, the system is… Continue reading GPS Localization with ROS, rviz and OSM

Visualizing Open Street Map Data with ROS & rviz

Localization and path planning are two of the most important components in autonomous robots. Robot Operating System provides a great foundation for working with maps and path planning. In this post, I will talk about how to use data provided by Open Street Map (OSM) and to visualize the osm data using rviz. I still… Continue reading Visualizing Open Street Map Data with ROS & rviz

Visualizing the Steering Model with Attention Maps

Introduction Convolutional neural networks are often known as "black boxes" for their mysterious nature. Unlike most programs that we write, computer scientist can't directly modify the content (weights) of the neural networks to improve their performance. In order to create better machine learning models, you can either do a heck more training or experiments with… Continue reading Visualizing the Steering Model with Attention Maps

A Gallery of our Self-Driving Cart

A picture is worth a thousand words Phase 1 Demo & Presentation Version 1.0.0 Exterior Testing on Feb 21st Testing on Jan 31st Other images Credits Thanks to the following people for taking the pictures: Ms. Richardson Mr. Hale (and his amazing editing) My parents Michael Meng   Feel free to reach out to me… Continue reading A Gallery of our Self-Driving Cart

The Limitations Of Our Deep Learning Powered Self-Driving Golf Cart

Introduction After our somewhat unsuccessful demo last Wednesday, my partner Michael Meng exclaimed, "there is no hope for deep learning". The future is not that grim, but Michael is right to a certain degree. Deep learning has flaws and our deep learning powered self-driving golf cart certainly has lots of flaws. Today, I would like to… Continue reading The Limitations Of Our Deep Learning Powered Self-Driving Golf Cart

Looking Into The Future

Yesterday, Michael and I had the final demo & presentation of our autonomous golf cart. Phase 1 of the development process is now completed, which means a couple of different things for us: For the rest of this trimester, we will no longer be developing the goal cart. We will continue the phase 2 of the… Continue reading Looking Into The Future

The Robustness of the Semantic Segmentation Network

Introduction On Feb 21st, Michael and I tested the autonomous steering and cruise control system. Unfortunately, the testing was largely unsuccessful. We encounter many issues with the system and testing conditions. This prompted me to think about the robustness of our systems, specifically the semantic segmentation system. Testing on Feb 21st. Thanks to Ms. Richardson Shadows Convolutional… Continue reading The Robustness of the Semantic Segmentation Network