Conceitos essenciais
Multi-residual Task Learning verbessert die Generalisierung von DRL-Steuerungen für kooperatives Eco-Driving.
Estatísticas
"Analyzing nearly 1200 traffic scenarios across 600 signalized intersections, we demonstrate that it emerges as a promising approach to synergize the strengths of DRL and conventional methods in generalizable control."
"Our MRTL policy improves benefits even in partial penetration scenarios when the nominal policy falls short."
"MRTL policies are more resilient to control noise, with only a 3% performance decrease compared to a significant 18% drop in the nominal policy."
Citações
"Multi-residual Task Learning verbessert die Generalisierung von DRL-Steuerungen für kooperatives Eco-Driving."
"Unsere MRTL-Politik verbessert die Vorteile auch in Szenarien mit teilweiser Penetration, wenn die nominale Politik nicht ausreicht."