
A population-based reinforcement learning (PBRL) framework that utilizes GPU-accelerated simulation to train robotic manipulation tasks by adaptively optimizing the set of hyperparameters during training, achieving superior performance compared to non-evolutionary baseline agents.


coremsg

scaling-population-based-reinforcement-learning-with-gpu-accelerated-simulation-for-efficient-exploration-and-hyperparameter-optimization


Scaling Population-Based Reinforcement Learning with GPU-Accelerated Simulation for Efficient Exploration and Hyperparameter Optimization


title_rewrite


A hierarchical reinforcement learning framework, Task-priority Intermediated Hierarchical Distributed Policies (TIHDP), enables multiple robots to adaptively transport objects of varying weights in environments with changing numbers of robots and objects.


task-priority-intermediated-hierarchical-distributed-policies-for-adaptive-multi-robot-cooperative-transport


Task-priority Intermediated Hierarchical Distributed Policies for Adaptive Multi-robot Cooperative Transport
