RRT-CoLearn: Towards Kinodynamic Planning without Numerical Trajectory Optimization @ICRA-cg8kk
RRT-CoLearn: Towards Kinodynamic Planning without Numerical Trajectory Optimization  @ICRA-cg8kk
Uploaded May 2018 | Updated November 2026, 2 days ago
ICRA 2018 Spotlight Video
Interactive Session Thu PM Pod I.8
Authors: Wolfslag, Wouter; Bharatheesha, Mukunda; Moerland, Thomas M; Wisse, Martijn
Title: RRT-CoLearn: Towards Kinodynamic Planning without Numerical Trajectory Optimization

Abstract:
Sampling-based kinodynamic planners, such as Rapidly-exploring Random Trees (RRTs), pose two fundamental challenges: computing a reliable (pseudo-)metric for the distance between two random nodes, and computing a steering input to connect the nodes. The core of these challenges is a Two Point Boundary Value Problem, known to be NP-hard. Recently, the distance metric has been approximated using supervised learning, reducing computation time drastically. Previous work on such learning RRTs use direct optimal control to generate the data for supervised learning. This paper proposes to use indirect optimal control instead, because it provides two benefits: it reduces the computational effort to generate the data, and it provides a low dimensional parametrization of the action space. The latter allows us to learn both the distance metric and the steering input. This eliminates the need for a local planner in learning RRTs. Experimental results on a pendulum swing up show 10-fold speed-up in both the offline data generation and the online planning time.
RRT-CoLearn: Towards Kinodynamic Planning without Numerical Trajectory OptimizationQ-CP: Learning Action Values for Cooperative PlanningSimplified Quasi-Steady Aeromechanic Model for Flapping-Wing Robots with Passively Rotating HingesAddressing Challenging Place Recognition Tasks Using Generative Adversarial NetworksAn Intelligent Control Scheme to Facilitate Abrupt Stopping on Self-Adjustable TreadmillsA Hierarchical Model for Action Recognition Based on Body PartsFireAnt: A Modular Robot with Full-Body Continuous DocksRobotic Cleaning through Dirt Rearrangement Planning with Learned Transition ModelsToward Soft Micro Bio Robots for Cellular and Chemical DeliveryIntent-Aware Multi-Agent Reinforcement LearningAutonomous Fixed-Wing Aerobatics: From Theory to FlightScrew-Powered Propulsion in Granular Media: An Experimental and Computational Study
ICRA 2018 |

RRT-CoLearn: Towards Kinodynamic Planning without Numerical Trajectory Optimization

SHARE TO X SHARE TO REDDIT SHARE TO FACEBOOK WALLPAPER