Handbook on Mobile and Ubiquitous Computing

Agent-Based Intrinsically Motivated Intelligent Environments

■

185

7.6.1 Unsupervised Learning Agent Model

The UL agent model used in this simulation is shown in Figure 7.17. The model is similar to the

MRL agent model presented in Section 7.5.1 but with several key differences: Firstly the motivation

process is absent. Secondly the learning process is disconnected from the sensation and action

processes. This is because learning is performed as a batch process driven by an implementation of

the WINEPI algorithm (Manilla et al. 1997) and the events generated by the sensation process are

used directly by the activation process without passing though