DJGrad [pdf]
The goal of this project is to create a network protocol capable of sending machine ML model updates between CAVs that are level 2-4 (human-in-the-loop), without the need for cloud communication or cellular networks.
We are interested in understanding the following aspects of our protocol:
- Networking: How well do model gradients spreadin realistic driving environments?
- See veins-dsrc folder
- Learning: Does our distributed learning protocol positively impact learning?
- See learning_sim folder
- Security & Privacy: How can a malicious vehicle attack other vehicles by exploiting this distributed learning protocol?
- See security folder
