This repository includes a AWD and RWD vehicle dynamics model as well as reinforcement learning framework for training a vehicle to get itself out of a ditch without any human intervention.
Repository
Levi Manring
8/25/21
This repository contains the modeling, simulation tools, reinforcement learning (RL) framework and results for multiple control simulations of a vehicle stuck in a ditch. The goal of this work is to provide some insight into vehicle automation and training a vehicle to intelligently escape from a ditch without human intervention.
BestResults -> contains the data files and plotting tools to quickly reproduce our results
RL_DDPG_AWD -> contains the RL framework using DDPG and assuming an AWD dynamic model with front and rear wheel drive and wheel slippage
RL_DDPG_RWD -> contains the RL framework using DDPG and assuming a RWD dynamic model with rear wheel drive and wheel slippage
RL_DDPG_RWD_NOSLIP -> contains the RL framework using DDPG and assuming a RWD dynamic model with rear wheel drive and assuming no wheel slippage
RL_PILCO_RWD -> contains the RL framework using PILCO and assuming a RWD dynamic model with rear wheel drive and assuming no wheel slippage
VehicleDynamicsModelAWD -> contains the AWD dynamics model
VehicleDynamicsModelRWD -> contains the RWD dynamics model
Each one of these files has its own README files for further explanation.