首页|Studies from University of Edinburgh Reveal New Findings on Androids (Online Mul ticontact Receding Horizon Planning Via Value Function Approximation)
Studies from University of Edinburgh Reveal New Findings on Androids (Online Mul ticontact Receding Horizon Planning Via Value Function Approximation)
扫码查看
点击上方二维码区域，可以放大扫码查看
原文链接
NETL
NSTL
By a News Reporter-Staff News Editor at Robotics & Machine Learning Daily News Daily News-Investigators publish new report on Ro botics - Androids. According to news reporting from Edinburgh, United Kingdom, b y NewsRx journalists, research stated, "Planning multicontact motions in a reced ing horizon fashion requires a value function to guide the planning with respect to the future, e.g., building momentum to traverse large obstacles. Traditional ly, the value function is approximated by computing trajectories in a prediction horizon (never executed) that foresees the future beyond the execution horizon. " Financial support for this research came from EU H2020 project Enhancing Healthc are with Assistive Robotic Mobile Manipulation HARMONY. The news correspondents obtained a quote from the research from the University o f Edinburgh, "However, given the nonconvex dynamics of multicontact motions, thi s approach is computationally expensive. To enable online receding horizon plann ing (RHP) of multicontact motions, we find efficient approximations of the value function. Specifically, we propose a trajectory-based and a learning-based appr oach. In the former, namely RHP with multiple levels of model fidelity, we appro ximate the value function by computing the prediction horizon with a convex rela xed model. In the latter, namely locally guided RHP, we learn an oracle to predi ct local objectives for locomotion tasks, and we use these local objectives to c onstruct local value functions for guiding a short-horizon RHP. We evaluate both approaches in simulation by planning centroidal trajectories of a humanoid robo t walking on moderate slopes, and on large slopes where the robot cannot maintai n static balance. Our results show that locally guided RHP achieves the best com putation efficiency (95%-98.6% cycles converge online )."
EdinburghUnited KingdomEuropeAndro idsEmerging TechnologiesMachine LearningRobotRoboticsUniversity of Edi nburgh
NETL
NSTL
