Graph-based Prediction and Planning Policy Network (GP3Net) for scalable self-driving in dynamic environments using Deep Reinforcement Learning.
Jayabrata ChowdhuryVenkataramanan ShivaramanSuresh SundaramP. B. SujitPublished in: CoRR (2023)
Keyphrases
- dynamic environments
- reinforcement learning
- reinforcement learning problems
- reinforcement learning algorithms
- action selection
- optimal policy
- plan execution
- single agent
- genetic programming
- belief space
- autonomous agents
- partially observable
- mobile robot
- reinforcement learning agents
- path planning
- potential field
- peer to peer
- heuristic search
- policy search
- changing environment
- markov decision problems
- markov decision process
- partially observable markov decision processes
- highly dynamic environments
- action space
- real environment
- state space
- agent systems
- control policy
- collision free
- reward function
- transfer learning
- planning problems
- multi agent systems
- function approximation
- agent based systems
- model free
- policy gradient
- dynamic programming
- reinforcement learning methods
- unmanned aerial vehicles
- semi supervised
- learning algorithm
- scalable video coding