# Possible extensions

Beyond the extension assignment, there exists a myriad of other possibilities. Many members of the Donkeycar discord and extended community participate in the process of technological exploration. Here are some possible directions for future projects building on top of CHI\@Edge Autonomous Vehicles:

1. Trade-offs of Running Inference Model versus Hosting Model on Chameleon Node
2. Comparison of Data Augmentations in Training
3. Attempting Offline Reinforcement Learning
4. Exploration of Methods to Actuate Bridging the Sim-to-Real Gap
5. Edge Computing Optimization for Inference Models

***

#### Required Knowledge for Extending

Many of these extensions are unlike the extension assignment for CHI\@Edge Autonomous cars. These projects require much more outside reading with answers that are not yet known. However, we can provide you with some reading that is helpful for pointing you in a direction. 

**READINGS:**

[Edge versus Cloud Computing Trade-Offs](https://ieeexplore.ieee.org/document/9006139) 

[Data Augmentation for Deep Learning](https://journalofbigdata.springeropen.com/articles/10.1186/s40537-019-0197-0)

[Zoo: Reinforcement Learning for Donkeycar](https://github.com/DLR-RM/rl-baselines3-zoo)

[Offline Reinforcement Learning: Bridging the Sim-to-Real Gap](https://arxiv.org/abs/2103.12021)

[Machine Learning Optimizations for the Edge](https://www.mdpi.com/1424-8220/23/3/1279)