social linkedin box blue 32
social facebook box blue 32
social twitter box blue 32
social facebook box blue 32
iit-advr-logo-v3
LabsHumanoids & Human Centred MechatronicsProjects ■ The kinematic Motion Primitives (kMPs)

The kinematic Motion Primitives (kMPs)

Humans, if compared to other animals, are able to accomplish a wider number of different tasks. In many cases, the motion produced is generated in real time to achieve a goal that has never been faced before. This flexibility is possible thanks to 1) their physical structure, that is the result of millennia of evolution, and to 2) their proficient neuromotor capabilities. This second point is fundamental to control a complex system like the human body.

The work on the kinematic Motion Primitives (kMPs) was an analysis aimed at understanding how humans control the complex motion of their whole body, and reproduce the human skills on the COmpliant huMANoid (COMAN) robot.

The first part of the project focused on locomotion. Five subjects were asked to perform a set of walking and running trials, both unconstrained (free arms motion) and constrained (holding an empty box, or a 5 kg load with both hands). The whole body motion of the subjects was recorded with a Vicon motion capture system at 250 Hz. The 34 joint trajectories were obtained from the cartesian trajectories of the 39 passive markers placed on their body (Figure 1). These joint trajectories were the basis of the analysis performed.

A Principal Component Analysis (PCA) was applied on each of the trajectories. What followed was the comparison between the main components returned by the PCA. It was noticed that the first 5 components, that together explained the 99% of joint trajectories variance, remained invariant among the different subjects (Figure 2), for different velocities of walking or running, or even other imposed constraints.

 

A statistical analysis confirmed what observed: all the different trajectories captured are accurately described by a small set of invariant signals, that we called kinematic Motion Primitives (kMPs).

From these kMPs a human-like center of mass trajectory was reconstructed and scaled down to the dimensions of the COMAN robot. This information, combined with valid engineered feet trajectories, was sufficient to generate human-like joint trajectories for our robot.

COMAN could perform a stable, highly dynamic, human-like walking (Figure 3), with knees straightening up to 5°, and a big vertical displacement of the COM, whose motion excited the springs in the actuators of COMAN more than the usual engineered walking trajectories do, exploiting then the intrinsic compliance of the robot to store and release energy at the proper phases of the gait.

 
Video: CoMAN (COmpliant huMANoid) human-like walking reconstructed from kinematic Motion Primitives (kMPs)

References

  • F.L. Moro, N.G. Tsagarakis and D.G. Caldwell
    A Human-like Walking for the COmpliant huMANoid COMAN based on CoM Trajectory Reconstruction from Kinematic Motion Primitives
    11th IEEE-RAS International Conference on Humanoid Robots, pp. 364-370, Bled, Slovenia, October 26-28, 2011

  • F.L. Moro, N.G. Tsagarakis and D.G. Caldwell
    Efficient Human-Like Walking for the COmpliant huMANoid COMAN based on Kinematic Motion Primitives (kMPs)
    IEEE International Conference on Robotics and Automation (ICRA), pp. 2007-2014, Saint Paul, Minnesota, USA, May 14-18, 2012

Last Updated on Thursday, 14 June 2012 10:20