Fig 6 : Comparison of succession rate of different approaches
We have illustrated the pathways of the navigation challenges, as shown in Figure 7, to illustrate the benefits of our approach more clearly. The graph clearly shows that our proposed method produces softer paths while the paths produced by the comparative methods oscillate to be different degrees.
Conclusion: We proposed a novel approach to control the telepresence robot during delayed signals by integrating LSTM with the DDPG model. It utilizes supervised and reinforcement learning to combine the indication and assessment signals. The proposed hybrid technique uses RNN in addition to the off-policy actor-critic architecture to identify the best dynamic treatments. The comprehensive experiments on the real-world manufactured telepresence robot generate a dataset by multiple traversing of the same path in a healthcare environment. The proposed approach showed appreciative results in simulation experiments compared to other methods. After the data generation, our proposed approach was used and revealed that the suggested method could boost controllability by up to 2.3% and offer more control during the lack of communication or commanding signals.