Figure 2 | Scientific Reports

Figure 2

From: Self-organizing neural network for reproducing human postural mode alternation through deep reinforcement learning

Figure 2

Screenshots of learning tracking-balancing tasks at motion tracking frequency f of target point in Eq. (5) with regard to 0.15 [Hz] and 1.5 [Hz]. Where, energy consumption penalization parameter \(\gamma\) in Eq. (6) is 20. The motion amplitude A of target point in Eq. (5) is 0.1 [m]. Ankle and hip stiffness remains 25 [Nm/rad] and 125 [Nm/rad].

Back to article page