MUR Driverless 2021 Team

Hi everyone! Similar to last year, we are a group of 6 students from the University of Melbourne, continuing the autonomous race car project. We’ve split into two groups; the autonomous perception and the autonomous control sub-teams. Keep a lookout for my teammates posts!

In the autonomous perception sub-team, we’ve divided the work into 4 areas:

1. Simultaneous localisation and mapping (SLAM)
2. Stereo vision
3. LiDAR
4. GPS/IMU i.e. localisation

In the autonomous control sub-team, we’ve divided the work into 3 areas:

1. Slow lap
2. Fast lap
3. Vehicle dynamics

Perception Timeline

March

• Obtain traffic cones and paint them blue and yellow to create the obstacle course.
• As soon as possible, we will install the baslar cameras, LiDAR, and Piksi (GPS/IMU sensor) on the MUR 2019E car to collect data.

March - June

• Validate the previous year’s work on the individual pipelines using the collected data or simulation through Gazebo as well as carefully thought out experiments.
  • Rigorous testing on the Jetson AGX Xavier.
• Use either the Husky robot or a shopping trolley to test the perception system as the 19E car will not be available.

July - September

Should the current perception system work, the plan for July to September is:

• Improve the individual pipelines (stereo vision, LiDAR, SLAM algorithm) to reduce the latency as much as possible.

October

• Test the upgraded perception system on the 21E car.
  • Collect more training data for validation.

Control Timeline

March

• Research into control concepts for concept review
• Understand the controller that is implemeted last year, find out points of improvement
• Design the autonomous control pipeline, integrating both slow and fast lap controller together

April - May

• Slow Lap Path Follower rewrite in C++
• Alumni (Dennis) finishes his implementation of MPCC
• Integrate MPCC into ROS
• Implement transition between slow lap and fast lap
• MPCC Controller tuning for fast lap

June - August

• Investigate physical testing on Husky
• Full autonomous pipeline integration with Perception team if possible, else implement autonomous control pipeline
• More simulations on race car and tunings if required

September - October

• Fine Tune Controllers
• Prepare for Endeavour Exhibition, hopefully physical demo with Husky with full autonomous pipeline

Secondary Goals

Throughout the course of the year, our secondary goals are:

• To build a robust testing environment such that any changes to individual pipeline can be easily tested within 24 hours. This involves improving:
  • The MUR Driverless Simulation (mursim)
  • The docker images
  • Reorganisation of the Github repository with clear labelling
• Create a solid documentation system such that future teams can easily pick up where we left off.