How does a human steer toward a visible goal object while avoiding obstacles in a cluttered environment? Psychophysics studies have shown that a goal acts as an attractor of heading, while obstacles act as repellors of heading. This neural modeling research shows how neural representations of heading, goal position, and obstacle position can generate dynamic steering behavior. The model includes network layers representing cortical areas MT, MST, and posterior parietal cortex (PPC). Area MT contains cells sensitive to speed and motion direction in local receptive fields. Area MST contains two subregions, dorsal MST (MSTd), which represents full-field motion, and ventral MST (MSTv), which represents object motion. Thus, heading is encoded by the population activity in MSTd, while goal and obstacle location are encoded by area MSTv. The output of these areas is combined to generate a dynamic steering signal.
View of the environment at the start of the simulation
The red pole is an obstacle and the blue pole is the goal
Illustration of the path taken by the model through the environment
