README.md

Inertial Odometry with Reinforcement Learning

The goal of the project is to perform end-to-end inertial navigation with deep reinforcement learning

Project To-Do List

Research

• Research and implement non-linear Kalman Filters
• Read and understand TLIO and IMO research papers

Implementation

• Implement and understand the concepts from the TLIO and IMO papers
• Create the Husky Dataset
• Recreate the TLIO Paper with the Husky Dataset

Simulation

• Create a simulated world environment
• Create and implement an end-to-end inertial navigation algorithm in the simulation

Real-World Application

• Zero-Shot Implementation
• Offline Reinforcement Learning
• Online Reinforcement Learning
• Imitation Learning (BC, Dagger,...)
• Classic Method (MPC,...)

Simulation To-Do List

Gazebo Classic

• Create Different Trajectories for the Husky
• Collect Simulation Dataset
• Create Reinforcement Learning Framework
• Train Online Reinforcement Learning

Isaac Sim

• Get Isaac ROS to work with Isaac Sim
• Implement VSLAM
• Create different different trajectories
• Collect Simulation Data
• Implement Online Reinforcement Learning