Welcome to the project report for Team 13! This page contains a report of all the progress we made throughout this busy and fun quarter, including our final project.
 |
 |
|---|
Team 13's assembled RC Car with the lidar placed at the front.
 |
 |
 |
 |
|---|---|---|---|
| Girish Krishnan [LinkedIn] | Muhammad Bintang Gemilang | Andy Zhang | Zhengyu (Van) Huang |
| Electrical Engineering | Mechanical Engineering | Electrical Engineering | Computer Engineering |
Our final project was themed around mapping an unknown environment. Our project involved the following tasks.
What we promised
- [✔] To implement SLAM (Simultaneous Localization and Mapping) using a lidar. This effectively creates a map of the environment around the robot, showing the locations of all objects present.
- [✔] To display the map generated from SLAM in real-time using a web application.
The challenges faced during the project were:
- Integrating the web application for live previewing (HTML/CSS/JS) with the Python code needed to run SLAM.
- Avoiding delays in the updating map.
The accomplishments of the project were:
- We were able to achieve a decent visualization that updates over time as the robot is driven around
- The visualization/map can be saved easily for tasks such as path planning.
Final Presentation
Weekly Update Presentations
Gantt Chart
- 3D Printing: Camera Mount, Jetson Nano Case, GPS (GNSS) Case.
- Laser Cutting: Base plate to mount electronics and other components.
Parts List
- Traxxas Chassis with steering servo and sensored brushless DC motor
- Jetson Nano
- WiFi adapter
- 64 GB Micro SD Card
- Adapter/reader for Micro SD Card
- Logitech F710 controller
- OAK-D Lite Camera
- LD06 Lidar
- VESC
- Anti-spark switch with power switch
- DC-DC Converter
- 4-cell LiPo battery
- Battery voltage checker/alarm
- DC Barrel Connector
- XT60, XT30, MR60 connectors
Additional Parts used for testing/debugging
- Car stand
- USB-C to USB-A cable
- Micro USB to USB cable
- 5V, 4A power supply for Jetson Nano
All measurements shown above are in millimeters (mm)
Our base plate was laser cut on a thick acrylic sheet. The circular hole at the end of the base plate is meant to hold the power on/off button. The long holes in the side of the plate are meant for wires to easily pass to and from the bottom of the plate.
 |
 |
|---|---|
| Camera Holder | Base for attachment to base plate |
The two parts of the camera mount shown above were screwed together. The angle of the camera was carefully chosen (facing downward approximately 10 degrees from the vertical) so that the road ahead is clearly visible. This is essential for accurate results in OpenCV/ROS2 autonomous laps.
One of our older camera mount designs is shown below.
 |
|---|
This camera mount consists of three parts: one base for attachment to the base plate, one middle piece to connect the base and the camera, and the camera holder. This camera mount design allows you to rotate the camera up and down. However, it is important that the rotating hinge is screwed securely so that the hinge doesn't wobble out while the robot does autonomous laps!
Credit to flyattack from Thingiverse, see: https://www.thingiverse.com/thing:3532828
 |
|---|
This case is excellent because it is robust and doesn't break easily, unlike most common Jetson Nano cases.
Note: some of these components and connections will vary depending on the exact components you have - check the component specifications carefully.
To install all the necessary Python modules needed, run the following on the Jetson Nano.
pip install -r requirements.txtFor our final project, we implemented a real-time visualization system for the Hector SLAM algorithm implemented using the lidar sensor. The base code for the SLAM algorithm is accessible in the Docker container provided to us in class, and the code for the real-time implementation is present in the slam_gui folder of this repository.
The SLAM real-time visualization GUI that we built has the following features:
- A web application whose routes are made using FastAPI in Python. Uvicorn is used to run the web server.
- HTML and JS to update the map in real-time
- The HTML and JS is interfaced with Python, ROS1, ROS2 and ROSBridge, so that the data collected is displayed on the web app.
- The interfacing process is difficult to implement directly in Python, so we use subprocessing to call relevant bash scripts that handle the processes in ROS1 and ROS2. These subprocesses are made to run in parallel using threading in Python.
To run the visualizer, first open up a docker container containing the ucsd_robocar ROS packages.
Run the following:
cd slam_gui
python slam_map.pyThis sets up the web app running on the Jetson Nano (although the app could potentially be run on any device, provided it can communicate with the Jetson Nano using the relevant ROS topics).
Opening up the web app on http://localhost:8000 reveals the GUI showing the results of SLAM in real-time. The code in slam_map.py can be adjusted to fine-tune the time-delay that occurs as the map updates.
Additional Scope for the Final Project
Although SLAM is useful for mapping an unknown environment, it can be useful to integrate GPS data with SLAM to provide better location accuracy. To implement this in Python, we created the folder gps_slam that contains starter code with lidar, PyVESC, and GPS implementation and a basic SLAM algorithm with the Kalman filter (implemented using the filterpy library in Python). This additional, nice-to-have part of the project hasn't been tested out yet, but we plan to get it working soon.
As part of the class deliverables and as preparation for the final project, here are our autonomous laps videos:
Donkey Sim
- Local Computer: https://youtu.be/lXEStSEVikQ
- GPU training: https://youtu.be/4_BzKP9-XAQ
- External Server: https://youtu.be/Yvo1yqRJhX4
Physical Robot
- DonkeyCar: https://youtu.be/bPUSS2g0Ves
- Lane detection using OpenCV + ROS2: https://youtu.be/omcDCBSrl2I
- Inner lane: https://youtu.be/9hN8HUlGcas
- Outer lane: https://youtu.be/nXZNPscVlX0
- GPS: https://youtu.be/Y3I9AWW1R6o
Thanks Prof. Jack Silberman and TAs Moises Lopez and Kishore Nukala for an awesome quarter! See you in DSC 178 next quarter, professor Jack ;)
-
Jetson Nano Case Design: https://www.thingiverse.com/thing:3532828
-
Lidar (LD06) Python Tutorial: https://github.com/henjin0/LIDAR_LD06_python_loder
-
SLAM tutorial, Dominic Nightingale. https://gitlab.com/ucsd_robocar/ucsd_robocar_nav1_pkg/-/tree/master/