The new algorithm allows robots to travel through the crowd.

According to CTOnews.com, October 13, robots have been widely used in takeout, express delivery, airport guidance, construction inspection and other fields, but one of the difficulties is to move among people.

At present, the robot mainly relies on cameras and other sensors to observe the surrounding environment, but this scheme is not stable, and the surrounding people will dynamically adjust the direction, so it is difficult to apply the robot in the crowd scene.

In his latest paper, Master of Science Chengmin Zhou proposed a reinforcement learning algorithm (RL) to guide robots to move among people.

This is a model-free reinforcement learning algorithm, which enables the robot to learn from historical experience, and after training or learning, the robot can travel even in challenging situations.

CTOnews.com note: this approach also has many challenges, such as slow learning, inability to analyze sensor information efficiently, resulting in inability to effectively deal with complex crowd scenarios.