Interested in Performing End-to-End RL Learning and experimentation? You Need DeepRacer

Source: Deep Learning on Medium

Interested in Performing End-to-End RL Learning and experimentation? You Need DeepRacer

Educational Autonomous Racing Platform for Experimentation with Sim2Real Reinforcement Learning

This research summary is just one of many that are distributed weekly on the AI scholar newsletter. To start receiving the weekly newsletter, sign up here.

Reinforcement Learning is an exciting AI technology with many applications for manipulation, locomotion, navigation, flight, interaction, motion planning and so on and so forth.

Training and evaluation for RL are complex and safety requirements and many go for is common simulation training. Additionally, one requires to have know-how in many areas including access to a physical robot, an accurate robot model for simulations, a distributed training mechanism and customizability of the training procedure such as modifying the neural network and the loss function or introducing noise. It is a challenging feat. To help deal with such challenges, researchers have introduced DeepRacer.

Educational Autonomous Racing Platform for Experimentation with Sim2Real Reinforcement Learning

To deal with current complexities in training and experimenting RL, researchers recently presented DeepRacer, an experimentation and educational platform for sim2real reinforcement learning. This is the first demonstration of model-free RL based sim2real at scale according to the researchers. It integrates state-of-the-art Deep RL algorithms, multiple simulation engines with OpenAI Gym interface, provides on-demand compute, distributed rollouts that facilitate domain randomization and robust evaluation in parallel.

Observation, action, and reward for DeepRacer agent

The researchers used the platform to demonstrate how a 1/18th scale car can learn to drive autonomously using RL with a monocular camera. Many users have simulated the model training and demonstrated sim2real RL navigation.

Potential Uses and Effects

While deep learning has boosted many advances in RL, we still have a long way to go to achieve robust systems. DeepRacer provides a platform that the machine learning community can use to take RL research and education to the next level. For instance, it helps minimize complexity and generalizability in training and evaluating RL tasks.

The platform can as well be used to systematically investigate the key challenges in developing intelligent control systems.

“DeepRacer is the first successful large-scale deployment of deep reinforcement learning on a robotic control agent that uses only raw camera images as observations and a model-free learning method to perform robust path planning,’’ say the researchers.

Open source code and video demo on GitHub

Read more: Platform for Experimentation with Sim2Real Reinforcement Learning

Thanks for reading, comment, share & let’s connect on Twitter, LinkedIn, and Facebook. Stay updated with the latest AI research developments, news, resources, tools, and more by subscribing to our weekly AI Scholar Newsletter for free! Subscribe here Remember to 👏 if you enjoyed this article. Cheers!