首页|Reports Outline Robotics and Automation Findings from Polytechnic University Tor ino (Transformer-based Model Predictive Control: Trajectory Optimization Via Seq uence Modeling)
Reports Outline Robotics and Automation Findings from Polytechnic University Tor ino (Transformer-based Model Predictive Control: Trajectory Optimization Via Seq uence Modeling)
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 DailyNews Daily News-A new study on Robotics - Robotics and Automation is now available. According tonews originating from Turin, Italy, b y NewsRx correspondents, research stated, "Model predictive control(MPC) has es tablished itself as the primary methodology for constrained control, enabling ge neral-purposerobot autonomy in diverse real-world scenarios. However, for most problems of interest, MPC relies on therecursive solution of highly non-convex trajectory optimization problems, leading to high computationalcomplexity and s trong dependency on initialization."Funders for this research include Blue Origin, NASA University Leadership Initia tive.Our news journalists obtained a quote from the research from Polytechnic Univers ity Torino, "In thiswork, we present a unified framework to combine the main st rengths of optimization-based and learningbasedmethods for MPC. Our approach e ntails embedding high-capacity, transformer-based neural networkmodels within t he optimization process for trajectory generation, whereby the transformer provi des a nearoptimalinitial guess, or target plan, to a non-convex optimization p roblem. Our experiments, performed insimulation and the real world onboard a fr ee flyer platform, demonstrate the capabilities of our frameworkto improve MPC convergence and runtime."
TurinItalyEuropeRobotics and Autom ationRoboticsPolytechnic University Torino
NETL
NSTL
