Last few points of linearly interpolated path are the same with fixed point density. Change point density to adaptive (based on distance between points), and validate points do not overlap

A TF master package/node should be created to store all the TF from sensors to chassis base_frame. Or other various transforms.

• Try tuning lanes v2
• Move to segmentation model - Check Apollo's reference

Creating an odometry node to facilitate dead-reckoning for the robot_localization package.

Needs front axle position transformation in control code so no more rear axle tracking.
Effect of rear axle tracking:
Matlab Sim Result with Rear Wheel Tracking

• Cross reference ENU converted back into latitude/longitude with http://vmddtech.com/ssp/latlong.html

• Change names of nodes to standardized naming convention
• Add pycache files etc to .gitignore
• Update setup shell script
• Add functional launch file
• CMakeLists and manifest
• Remove unused code
• Start README

Needs to debug and design solution for the demo failure.

Aligning the magnetometer orientation with the converted GPS X-Y Position Coordinates.
Finding the angular rotation required to match magnetometer coordinate system with global coordinate system
Running the car straight in any direction and gathering both magnetometer and GPS data
Figure out the difference in angle between the two and subtract

Note that before running rosserial you must install the binaries, follow steps 2.1.1.

In order to communicate to Arduino you will also need to install jetsonhacks/installACMModule (guide).

Create person detector with rough distance calculation.

Subscribes to:

• raw image input

Publishes:

• x location of the center of each bounding box relative to center of screen
• distance (in meters) to each person

Fix GPF Bug
Verify everything is working as it should

It is not being used

Currently points to the localization node. Should directly pull orientation from imu node instead.

• Fix README
• Fix or replace startup shell script
• Update CMakeLists and Package.xml documentation
• Update .gitignore
• Remove unnecessary hidden files

To interface all reference frames. Just start with IMU conversion and output

Integrate all current perception libraries into our main stack.

A transformation of 105 degree was made to account for magnetic declination. However, collected data was shown that the IMU package might have already accounted for the 15 degrees of declination.

The noisy data is the estimation of the heading value using backward differencing of the position data and took the arctangent of the dx and dy. The data has shown identical values to the 90 degree transform which is shown in yellow, not the real time IMU heading (105 degree transform) in light blue.

Begin working and committing low-level control Arduino code to the /arduino directory.

Backup orientation feedback, in case IMU and GPS frame matching don't happen on time.

Figure out horizontal direction camera detection relationship to Lidar angles

Utilize camera people detection and reinforcement layer for Lidar object detection

Integrate all current localization code into main stack.

Please create all the gateway module as a catkin package under /src/utility/interface/*

A development guideline for coding style, git good practices, and ROS good practices shall be created.

• Choose in parameter for path_interpolation for conversion type
• Start launch file for TF

Edit "sh ext_package_build.sh" script to install libuvc.

Without this install, the following error occurs:

The ros node initializes and was able to start but there was no GPS tracking coordinate publishing.

sparkfun/Arduino_Boards#32

This fix the issue. The error would show a path to the bossac file. Just copy the file path to the command and it should work.

1. Set up sensors, vehicle, and world TF.
2. Set up odom information
3. Figure out a mapping solution

Make branch for TX2 use only. Remove all unnecessary files to lightweight TX2 processing

Timer function was acting up and wouldn't fire the callback at all. Probably due to the class structure it is in.
Performed a hotfix offline by moving the control command computation and publishing into path callback function.

Slope matching function was returning NaN due to projected slope does not match interpolated slope. The hotfix was bypassing the matching check, which might be a calculation flaw in the code or the logic.

Intermediate node between steering/throttle/braking input to steering/throttle/braking CAN Messages.

https://docs.google.com/document/d/1-gH8Ibcog-r76q7y4_Z1wQ6STw4litTK_eBTXs2pq0g/edit

Ensure sign convention checks out.

High-level control command output will follow the convention below:

• Vehicle Turning left (steering wheel Counter-Clockwise) is a POSITIVE quantity.

GPF isn't quite working as well as desired so the parameters must be tweaked to achieve better results.

• Finish exporting clusters as pcl::PointCloud2 and std::vector<pcl::PointIndices>
• Attempt to figure out strange gpf failing (suspension?)
• Implement move semantics to decrease Vec3 copying

Provide enough information for all team members to easily utilize package

• different parsing modes for manual input path
• Operating modes for conversions

edit driver

• Write Node file for kalman filter which handles measurement synchronization
• Edit kalman filter framework to allow for dynamic sensor noise covariance

http://wiki.ros.org/razor_imu_9dof

Refer to this documentation.