Open Sourced AV Software

A repository to host the open source projects for Autonomous Driving research group.

Community Creator: Mayur Waghchoure and friends

RULES:

• All the Members are encouraged to equally contribute to the projects.
• All the members are encouraged to come forward and share their knowledge and materials to everyone in the group with the motto of teach by learning.
• Feel free to ask questions if one is stuck at any point for smooth progress of the project
• All the members(intermediate level) of the group with experience are urged to spare some time to answer the query.
• Follow the format decided so that everyone will be at same pace.
• All the members are requested to actively participate in the discussion.
• Imp: Any member if found frequently not attending the meeting or not contributing to the project may loose their opportunity to work on the project

Project Timeline:

Target 1: to be achieved before 5th May

1. Literature Review -
2. Deciding on the final architecture - To be Finalized on 4th May(Saturday) before having meeting with Mayur on 5th May
3. Prepare a Small ppt summarising the progress

Target 2: 02 June 2024

1. Methodology and Implementation 70% completion

Target 3: 30 June 2024

1. Complete Implementation and Results with final presentation

PROJECT PLAN:

Detailed map of all the elements team needs to accomplish to reach project’s goals

Target 1

Initiation and Planning	Timeline	Status
Team Formation	14/04/2024	Done
Git fork and clone repository	< 05/05/24	In Progress
Reading Research Papers	22/04/24> Week2 <26/04/24	In Progress
Literature Review from Papers	29/04/24> Week3 <3/05/24
Developing Abstract and Architecture	29/04/24> Week3 <3/05/24
Preparing PPTs	29/04/24> Week3 <3/05/24

Target 2

Execute and complete task 06/05/24 - 31/05/24

Target 3

Towards Project closure 03/06/24 - 25/06/24

Projects

I am hoping all our projects would fit together as a jigsaw puzzle to get the autonomous driving capable software

Track A

First visual/sensor fusion + Semantic BEV + detection + tracking + motion prediction to get the current tracks and future trajectories

Track B

Semantic SLAM, 3D mapping, and Localization or point cloud project to localize the object and to get the map

Track C

Then path planning would jump in and then controls

Track D

DBW, Motor controls and Simulation based testing

Projects

Track A Projects:

1. Visual Fusion/ Sensor fusion

To fuse the data from multiple modalities to get the understanding of the scene. e.g. Lidar and camera fusion to get the list of detected objects

2. Semantic BEV

To merge the data from the multiple cameras toget the birds eye view of the scene, and then perfrom the downstream tasks on it. first try this with IPM , and then using the transformers.

3. Object detection and tracking

First try out one stage and two stage object detector on camera images to detect the vehicles, pedestrians, traffic lights and Signs, and track them further Second, to perform this on Birds eye view(BEV) maps of the scene by merging info from multiple cameras

4. Motion prediction

To perfrom the track prdection using the model based approach followed by hybrid approach and then deep learning approach for the long term motion prediction

5. Sensor fusion

Fuse the data from different sensors using the different data association methods and kalman filters

Track B projects:

6. Localization

1. GPS and Odometry based Localization using the Kalman filters
2. Visual odometry
3. HD Map based Localization a. Lanelet maps b. Lidar Point Cloud based Localization
4. Mapless Localization

please find the relevant research papers, and try to implement first GPS +IMU based Localization. Eventually try the HD Map based localization

7. 3D mapping and 3D scene reconstruction

To create the map of the environment using the lidar or camera data.

8. Semantic SLAM

using the Graph neural networks create the map of the environment and localize inside it

Track C projects:

9. Path Planning

we would want to develop the path planning algorithm for the autonomous vehicles.

1. Global planner: here we would want to develop the global planner using the lanelet maps or Google maps api
2. Behavior planner: it would be rule based in the beginning using the state machines, which we can later do using the RL
3. Local planner: local planner would be of different style, which would include the path profiling and velocity profile planning, some algo are state lattice planners, RRT* etc. We could also develop MPC controller do the trajectory planning and tracking

10. MPC based trajectory planning and control to enable the connected mobility

11. Longitudinal and Lateral control algorithms

Track D projects

12. Drive by wire module

13. Motor controllers

14. Communication network

15. Simulation for the testing

16. ADAS feature development using the Model based design methodology