Page 62 - Read Online

Harib et al. Intell Robot 2022;2(1):37-71 https://dx.doi.org/10.20517/ir.2021.19 Page 57

Table 4. NN-based control in robotic manipulation - an overview
Approach Employed by…
Backpropagation Elsley [98] (1988), Huan et al. [109] (1988), Karakasoglu and Sundareshan [100] (1990) and Wang and Yeh [110] (1990)
[103-108]
CMAC learning Miller et al. (1987-1990)
[109] [139]
Adaptive NNs/PG Huan et al. (1988) and He et al. (2017)
table
[133] [128] [131,132] [134]
NNs for flexible joints Hui et al. (2002), Gueaieb et al. (2003), Chaoui et al. (2004), Subudhi and Morris (2006),
[130] [126] [140] [142]
Chaoui et al. (2006), Chaoui and Gueaieb (2008), He et al. (2017) and Sun et al. (2017)
[136] [137] [138]
NNs for multiple arms Hou et al. (2010), Li and Su (2013) and Li et al. (2014)
Feedforward and Chaoui et al. [135] (2009)
feedback
RNNs
[101]
Hopfield net Xu et al. (1990)
Comparison Wilhelmsen and Cotter [102] (1990)
NNs: Neural Networks; CMAC: cerebellar model articulation controller; RNNs: recurrent NNs.

Figure 8. Classification of AI categories.

objectives. It combines function approximation and goal optimization to map states and actions to the
rewards they result in. The combination of NN with RL algorithms led to the creation of astounding
breakthroughs like Deepmind’s AlphaGo, an algorithm that beat the world champions of the Go board
game .
[147]

As mentioned earlier, RL is a powerful technique for achieving optimal control in robotic systems.
Traditional optimal control has the drawback of requiring complete understanding of the system’s
dynamics. Furthermore, because the design is often done offline, it is unable to deal with the changing
dynamics of a system during operation, such as service robots that must execute a variety of duties in an
unstructured and dynamic environment. The first chapter of this paper has shown that adaptive control, on