An integrative computational approach for obstacle avoidance during action selection

Area of Science:

• Computational neuroscience
• Motor control
• Cognitive robotics

Background:

• Action selection in cluttered environments is a complex cognitive and motor challenge.
• Existing models often struggle to integrate goal-directed behavior with obstacle avoidance.
• Understanding how the brain balances approach and avoidance is crucial for naturalistic movement.

Purpose of the Study:

• To develop a computational model for action selection that integrates target approach and obstacle avoidance.
• To extend stochastic optimal control theory with a novel obstacle avoidance framework.
• To provide a unified measure of "relative desirability" for dynamic policy integration.

Main Methods:

• Proposed a computational model extending stochastic optimal control theory.
• Decomposed action selection into weighted combinations of approach and avoidance control policies.
• Integrated value information from goals, obstacles, and actions into a unified desirability measure.
• Evaluated the framework using simulated target-reaching tasks in cluttered environments.

Main Results:

• The model successfully captured key features of human motor behavior in simulated cluttered environments.
• Demonstrated the influence of obstacle properties on movement trajectories.
• Observed the model's ability to replicate the transient tendency to move towards obstacles before avoidance.

Conclusions:

• The proposed framework offers a novel approach to understanding action selection in complex, naturalistic settings.
• Provides insights into the dynamic interplay between approach and avoidance behaviors.
• The model serves as a comprehensive framework for studying action selection challenges faced by the brain.