Biped locomotion control through a biologically-inspired closed-loop controller

The aim of this project is to study the applicability of a bio-inspired controller for rehabilitation, either for assistance and/or therapy purposes.

Currently, as the present walking orthotic/ prosthetic systems are not sufficiently prepared to successfully react to unexpected real-world environment changes (such as uneven ground, slopes, obstacles, pushes, . . . ), the ultimate goal is to include a CPG model provided of the principles underlying the robust control of locomotion, the rules and the degree of pre-programmed behavior that may offer the flexibility to adapt to changes in the environment.

This study has great application in the project of autonomous robots and in the rehabilitation technology, not only in the project of prostheses and orthoses, but also in the searching of procedures that help to recuperate motor functions of human beings.

Several challenges are included in the biped locomotion field:

  • Understand the underlying the neural control principles of locomotion;
  • Exploration of the braistem - spinal - biomechanics interaction;
  • Capability of the neural drive of the biped system to generate its own walking gaits
  • Adressing of stability issues due to posture balance and controller-biomechanics coupling problems;
  • Generation and control of adaptive biped locomotion against unexpected disturbances and environmental variations.

The referred problems are being tackled through a biomimetic architecture, more specifically the modelling of Central Pattern Generators (CPGs) using adaptive nonlinear oscillators (AFOs).

Locomotion behavior can be successfully achieved through the effective reciprocal and dynamic coupling between the brain, body and environment, by means of internal and external feedback pathways. Endorsing this line of thought, the overall system design proposed is thus composed of three main blocks, namely the High-level control system (represented by the brainstem or supraspinal sites), the Low-level control system (represented by the spinal cord CPGs) and the biomechanical Biped walker system. A schematic of their interaction is shown in Figure 1.

A biped robot was employed in computer simulation through a biped walker simulator. The body consists of a two-dimensional, rigid five-link system in the sagittal plane including a torso and two identical legs. In Figures 2 and 3 are depicted the mechnism structure of the biped and the GUI to simulate the biped walking.

Concerning the oscillators' properties, a network of AFOs is capable of learning an arbitrary (quasi-periodic) signal through the learning adaptive concept. Such a mechanism can be potentially beneficial for adapting the intrinsic frequencies of the oscillators to the frequency components of sensory feedback signals, for instance, from a mechanical system and therefore replicating the sensory signals. By way of example, a specific DOF trajectory is learned from the biomechanical system in Figure 4.

Below in attachment there are available some videos with the aim of demontrating:

  • The importance of the phase modulation of the oscillators to enhance the coupling and synchronization between the controller and the mechanical system;
  • The importance of internal modulation on the production of non-periodic, voluntary movements;
  • Adaptive walking to environmental changes such as tilted ground transtions or uneven ground with obstacles.
People involved in this project: 
Related publications: 
Matos, V., C. P. Santos, and C. M. A. Pinto, "A brainstem-like modulation approach for gait transition in a quadruped robot", Intelligent Robots and Systems, 2009. IROS 2009. IEEE/RSJ International Conference on, pp. 2665 -2670, oct., 2009.
Project status: 
Project in progress
Importance of phase modulation on tilted ground.avi1.44 MB
Walking on downslope and upslope terrains.avi1.45 MB
Walking on upslope and downslope terrains.avi1.58 MB
Obstacle avoidance through internal modulation.wmv8.89 MB