Models of gaze control for manipulation tasks

Abstract

Human studies have shown that gaze shifts are mostly driven by the current task demands. In manipulation tasks, gaze leads action to the next manipulation target. One explanation is that fixations gather information about task relevant properties, where task relevance is signalled by reward. This work presents new computational models of gaze shifting, where the agent imagines ahead in time the informational effects of possible gaze fixations. Building on our previous work, the contributions of this article are: (i) the presentation of two new gaze control models, (ii) comparison of their performance to our previous model, (iii) results showing the fit of all these models to previously published human data, and (iv) integration of a visual search process. The first new model selects the gaze that most reduces positional uncertainty of landmarks (Unc), and the second maximises expected rewards by reducing positional uncertainty (RU). Our previous approach maximises the expected gain in cumulative reward by reducing positional uncertainty (RUG). In experiment ii the models are tested on a simulated humanoid robot performing a manipulation task, and each model's performance is characterised by varying three environmental variables. This experiment provides evidence that the RUG model has the best overall performance. In experiment iii, we compare the hand-eye coordination timings of the models in a robot simulation to those obtained from human data. This provides evidence that only the models that incorporate both uncertainty and reward (RU and RUG) match human data.

Modeling Task Effects in Human Reading with Neural Network-based Attention

“…The models of Reichle and Laurent (2006) and Lewis et al (2013) are particularly related to ours in that they optimize policies for rewards that explicitly trade off economy with accuracy (on word identification, Reichle & Laurent, 2006, or lexical decision, Lewis et al, 2013. Beyond language, models of visual behavior based on reinforcement learning have been proposed in other domains (e.g., Acharya, Chen, Myers, Lewis, & Howes, 2017;Butko & Movellan, 2008;Hayhoe & Ballard, 2014;Nuñez-Varela & Wyatt, 2013;Sprague, Ballard, & Robinson, 2007), using both policy-gradient methods like in our model (Butko & Movellan, 2008) and Q-Learning algorithms (Acharya et al, 2017;Nuñez-Varela & Wyatt, 2013;Sprague et al, 2007).…”

Parameter Inference for Computational Cognitive Models with Approximate Bayesian Computation

“…Computational rational agents have been used to model a number of phenomena in HCI (Payne & Howes, 2013). Applications relevant to this paper include menu interaction (Chen et al, 2015), visual search (Hayhoe & Ballard, 2014;Myers, Lewis, & Howes, 2013;Nunez-Varela & Wyatt, 2013;Tseng & Howes, 2015), and decision-making (Chen, Starke, Baber, & Howes, 2017).…”

Inferring Cognitive Models from Data using Approximate Bayesian Computation

“…Computational rational agents have been used to model a number of phenomena in HCI [36]. Applications relevant to this paper include menu interaction [13] and visual search [20,30,34,44].…”