Here is a dual-agent reinforcement learning framework for robots that puts whole-body control in 2 agents. The lower body agent ensures stable movement under external force while the upper body agent tracks end-effector positions. These agents are trained jointly in simulation. As the researchers explain:
FALCON achieves 2x more accurate upper-body joint tracking, while maintaining robust locomotion under force disturbances and achieving faster training convergence. Moreover, FALCON enables policy training without embodiment-specific reward or curriculum tuning.
FALCON achieves 2 times more accurate upper-body joint tracking. You can find out more here.
[HT] [credit: Yuanhang Zhang Yifu Yuan Prajwal Gurunath Tairan He Shayegan Omidshafiei Ali-akbar Agha-mohammadi Marcell Vazquez-Chanlatte Liam Pedersen Guanya Shi]
