Robot body self-modeling algorithm: a collision-free motion planning approach for humanoids

Area of Science:

• Robotics
• Artificial Intelligence
• Biomechanics

Background:

• Humanoid robot motion planning faces challenges with redundancy, stability, feasibility, and collision avoidance.
• Human nervous system strategies offer inspiration for robot movement control.
• Self-modeling, inspired by human body awareness, is a key concept.

Purpose of the Study:

• To integrate self-modeling into an optimal motion planning framework.
• To detect and prevent collisions between a manipulated object and the humanoid robot.
• To ensure stable and feasible dynamic motion for humanoid robots.

Main Methods:

• Developed twelve parametric functions as self-models to define the humanoid robot's body boundaries.
• Utilized these self-models to compute a safe region for object manipulation, preventing collisions.
• Employed four distinct objective functions in motion simulations to test algorithm robustness.

Main Results:

• Successfully demonstrated collision avoidance between the manipulated object and the humanoid robot.
• Validated the reality and stability of the planned robot motions under various dynamics.
• Confirmed the effectiveness of the self-modeling approach in optimal motion planning.

Conclusions:

• Self-modeling is an effective strategy for enhancing humanoid robot motion planning.
• The proposed framework successfully addresses dynamic collision avoidance and ensures motion stability.
• This research contributes to more capable and safer humanoid robot operation.