Inverse optimal control to model human trajectories during locomotion

Area of Science:

• Robotics
• Biomechanics
• Human-Robot Interaction

Background:

• Cobotic applications necessitate understanding human behavior for seamless interaction.
• Humanoid robots require accurate models of human movement for tasks like co-navigation and co-manipulation.

Purpose of the Study:

• To model human Center of Mass (CoM) trajectories during locomotion.
• To generate human-like trajectories using an optimal control scheme.
• To propose a metric for assessing model accuracy against human behavior.

Main Methods:

• Recorded and analyzed CoM trajectories of 10 healthy subjects.
• Employed inverse optimal control to identify the optimal cost function.
• Generated human-like trajectories and compared them to recorded data.

Main Results:

• The developed model provides an accurate approximation of average human walking trajectories.
• Despite natural variability in human gaits, the model captures essential movement patterns.
• The model's accuracy is sufficient for improving cobotic task performance.

Conclusions:

• The study successfully models human locomotion CoM paths for robotic applications.
• The generated trajectories enable humanoid robots to anticipate human movements.
• This research advances the fluidity and safety of human-robot collaborative tasks.