Self-Segmentation of Sequences algorithm using eligibility traces in POMDP environments

This paper presents a new hierarchical reinforcement learning (RL) algorithm to speed up learning, to make it more robust to hidden states, and to handle non-deterministic problems. This on-line RL algorithm is called SSS(/spl lambda/), which extended Sun and Sessions´s original Self-Segmentation of Sequences (SSS) algorithm using eligibility traces. SSS(/spl lambda/) is compared with the original SSS for partially observable navigation tasks. The results of extensive simulations demonstrate that SSS(/spl lambda/) is clearly outperforming the original SSS.