nova-utd / navigator Goto Github PK

Navigator, our self-driving vehicle software stack

Home Page: https://nova-utd.github.io/navigator

License: Other

Shell 0.31% Dockerfile 0.05% Python 7.30% CMake 1.65% C++ 10.91% C 76.48% HTML 0.03% JavaScript 0.22% Perl 0.81% Makefile 0.09% Java 1.57% GAP 0.28% ANTLR 0.18% StringTemplate 0.11%

autonomous-driving autonomous-vehicles robotics ros ros2 self-driving

navigator's Introduction

Navigator is a simple, modular, open-source autonomous driving stack. It's developed by Nova, an undergraduate research team at UT Dallas.

For docs, visit our documentation site.

For info about our group, Nova, visit our group site.

navigator's People

Contributors

Stargazers

Watchers

Forkers

forrest-z bytetok ignss autodrivesoftware jpahm b-anjana ayush-sharma-py chtun akevinge dav1dddddddd

navigator's Issues

Calculate desired yaw rate

Implement traffic sign classifier

Generate splines for map data

Test classifier

Write wiki article describing outputs of ZED wrapper

Generate position paths

Calculate desired yaw rate

Create list of vehicle parameters

Determine miscellaneous constants

Path Planner: Assign costs to paths

Add safety features to BPC

Decide expected input of trajectory

Determine cost threshold for safety

Find suitable 3D pedestrian/vehicle classification model

Pass perception data as input to state machine

Path Planner: Generate paths

BP: Link State Machine with Perception & Path data

Pass path/route data as input to state machine

Transform map and localization information into a usable format

Control: Develop adaptation algorithm

Tweak route planner to match needs

Determine interface between path and behavior planner

Create initial diagram of all accepting states and transition functions

Connect state machine with path planner

Ensure that vehicle has a state and transition for every possible scenario of vehicle

Accept commands from the Safety Manager

Determine existing problems with localizer

Modify PID controller for throttle and brake

Problem

Our lateral (steering) controller is currently just a heuristic, and might not adapt well to different driving environments.

Our longitudinal (throttle and brake) controller does not exist yet, except for the trivial "bang bang" controller for SVL.

Solution

Apply adaptive PID control techniques to control both the steering, throttle, and brake.

Success conditions

Two nodes exist. One publishes steering commands on a /control/steering topic, and the other publishes on both a /control/throttle and /control/brake topic.
Right now, Navigator controls steering in two stages. The first generates a desired steering angle, and another (PID) controller converts this into effort. These should be merged so that the steering effort is generated directly.
Both controllers can (theoretically) adapt to different driving conditions, such as fairly sharp turns, wet roads, etc.
Both controllers are tested and validated in tandem using the SVL simulator.

Determine weights for different cost functions

"unable to resolve host vde: Temporary failure in name resolution"

When executing sudo in fresh ADE installation, the terminal hangs for a few seconds and then fails:

docker@vde:~$ sudo true
sudo: unable to resolve host vde: Temporary failure in name resolution

Reverting to 7214008 fixes it... Must be something wonky with the host configuration in Docker.

Clone the wrapper.
Download the dependencies.
Compile them in a ROS environment and see if it runs.
Test if the camera can be accessed through the wrapper.
Test to see if basic features advertised, such as object detection, are working as intended.
Make sure the data is being published to the correct topics.
Determine optimal wrapper configuration
Add wrapper code to main.launch.py