Article Figures & Data
Figures
- Figure 1.
Schema of the neurorobot equipped with an artificial eye, provided with a 7 × 7 light receptor retina and controlled by two pairs of simulated extraocular muscles.
- Figure 2.
A–C, Different implementations of the attentional networks with interhemispheric inhibition (Koch et al., 2011) and DAN/VAN architecture (Corbetta and Shulman, 2002). D, E, Two implementations of right-hemisphere networks with bilateral competence (Heilman and Van Den Abell, 1980; Mesulam, 1981) and no interhemispheric inhibition. Arrows indicate connections that can be either excitatory or inhibitory; red connections with triangular arrowheads denote excitatory connections; blue round arrowheads represent inhibitory connections. LH, left hemisphere; RH, right hemisphere; Canc., cancellation units; LDAN and RDAN, dorsal attention networks in the left and in the right hemisphere, respectively; LVAN and RVAN, ventral attention networks in the left and in the right hemisphere; LVF and RVF, left and right visual field. Right and left VANs have the same number of neurons but different patterns of connection strength.
- Figure 3.
Mean percentage of correct cancellations computed across 30 trials for each population of 40 neurorobots provided with neurocontrollers A–E. The middle bar of the boxplot indicates the median of the tested population. The top and the bottom of the box indicate respectively the first (q1) and the third (q3) quartiles. Whisker length extends until the last data point that is not considered as an outlier, i.e., a point that is greater than q3 + 1.5 × (q3 – q1) or less than q1 – 1.5 × (q3 – q1). There were no outliers in the current dataset.
- Figure 4.
Average x values of the first cancelled target, computed across 30 trials for each population of 40 neurorobots provided with neurocontrollers A–E.
- Figure 5.
Average x values of the first cancelled targets, for all the neurorobots provided with neurocontrollers C, C0, and C1. Only average x values of neurocontrollers C and C1 significantly differed from 0.
- Figure 6.
Average position on the x-axis of the first cancelled targets for human participants (H) and for artificial neurorobots equipped with neural networks A, B, C, C0, C1, D, and E.
- Figure 7.
Laterality indexes of DAN activation computed for individuals equipped with neurocontroller C and C0. A value of 0 means that activation in left and right hemisphere DANs is balanced; positive values denote prevalence of right hemisphere DAN, negative values indicate prevalence of left hemisphere DAN.
- Figure 8.
Average activation of hidden neurons in right hemisphere DAN (RDAN) and in left hemisphere DAN (LDAN), for the first 30 steps of individuals equipped with neurocontrollers C and C0. The activity scale goes from 0 (black) to 1 (white). Note the early, large left-right asymmetry in neurobiologically inspired C agents, which subsequently decreases. The symmetrical C0 agents do not show any asymmetries of performance.
- Figure 9.
Relative frequencies of the distribution of the position of the first cancelled target for 101 human participants (see experiment 1) and for the populations of neurorobots C (equipped with the biologically inspired neurocontroller), C0 (presenting symmetrical DANs), and C1 (with VAN-DAN connections only present in the right hemisphere).
- Figure 10.
Coordinates of canceled targets as a function of the temporal sequence of cancellation in human participants and in neurorobot population C. Error bars represent credible intervals of 95%.
Extended Data
- Extended Data 1
Neurobots and cancellation task have been implemented on the basis of Evorobot*, an open source tool for running evolutionary experiments (http://laral.istc.cnr.it/evorobotstar/; Nolfi and Gigliotta, 2010). The present code, written in C++, includes header (.h) and source (.cpp) files of the modified version of Evorobot*. In particular, the motor control is defined in file epuck_sm.cpp (function move_robot retinaMotorControlType 20); neurorobots and task are initialized in file epuck.cpp (functions: initialize_robot_cancellationTask3(); initialize_world_cancellationTask; update_sensors; update_motors). The fitness function is defined in fuction ffitness_cancellationTask3. Download Extended Data 1, ZIP file.
In this issue
