Modelling 3D saccade generation by feedforward optimal control

Area of Science:

• Computational Neuroscience
• Ocular Motor Control
• Robotics

Background:

• The human saccadic eye-movement system faces the degrees-of-freedom problem, requiring three rotations but having only two degrees of freedom for gaze control.
• Eye movements during saccades typically adhere to Listing's law (LL), where cyclotorsion is zero, and follow nonlinear main-sequence dynamics.

Purpose of the Study:

• To explain the emergence of Listing's law and saccade main-sequence dynamics using a computational model.
• To investigate how optimal control principles applied to a simplified eye model can reproduce observed saccadic eye movement properties.

Main Methods:

• Developed a 3-degree-of-freedom robotic eye model with antagonistic muscle pairs and independent motor commands.
• Employed numerical linearization and system identification for controller development.
• Applied optimal control principles to the simulated model to reproduce saccadic behavior.

Main Results:

• The optimal control model successfully reproduced Listing's law and the saccade main-sequence dynamics.
• Identified four key terms in the cost functional: energy expenditure, movement duration, gaze accuracy, and static muscle force.

Conclusions:

• Listing's law and saccade dynamics may emerge from a common mechanism optimizing speed-accuracy trade-offs.
• Minimizing total muscle force during fixation is also a crucial factor in generating realistic saccadic eye movements.