Priority based movement control improves artificial agent performance
From Visual Search to Movement Control: A Priority Field for Artificial Agents
Robotics
Summary
Humans pay attention to important things around them by combining what stands out visually, their current goals, and past experiences. The authors created a simple model that mimics how humans move their eyes when searching visually, capturing key behaviors like ignoring distractions and using memory of past targets. They then extended this idea to control artificial agents reaching goals while avoiding moving obstacles, showing that this priority-based approach helps agents learn faster and handle new challenges better. Adding a memory feature also allowed the agents to predict where future goals might appear, similar to humans.
What this means in practice
- •For robotics engineers: Develop robots that navigate dynamic environments more efficiently by using priority-based fields to decide where to move next.
- •For autonomous vehicle developers: Improve obstacle avoidance and goal reaching in self-driving cars by integrating priority-based attention models trained from simple demonstrations.
- •For video game ai developers: Create game characters that move realistically by anticipating goals and avoiding obstacles using priority maps with memory effects.$Commercial implications: Enables more adaptive and human-like AI movement in games, enhancing user experience and immersion.
Authors
Han Zhang, Zhong Cao
Abstract
Human spatial attention is widely conceptualized as being guided by a priority map that integrates perceptual salience, current goals, and past experiences. Here, we extend priority-based computation to movement control in artificial agents. We first introduce a lightweight model of visual search based on an integrated priority map. Trained on human saccades, it reproduced key behavioral patterns, including oculomotor suppression and history-driven selection. Extending the search model, we equipped an artificial agent with a priority field and evaluated its performance in a reach-avoid task that required reaching a goal destination while avoiding moving obstacles. Compared with alternative architectures, priority-field agents trained more efficiently and performed better in unseen, complex scenarios, even from simple demonstrations. Adding a simple memory mechanism also produced human-like, history-driven effects in anticipating the likely location of the upcoming goal. These findings suggest that priority-based computation may provide a promising foundation for movement control in artificial agents.