Behavioral performances of our legged robots are still far behind those of biological systems. Energy efficiency and locomotion velocity of our robots, for example, are orders of magnitude lower than those of animals, and in order to fill the gap, it requires a radically new approach in the design and control processes. From this perspective, we have been exploring a novel approach to design and control of legged robots which makes use of free vibration of elastic curved beams. We found that this approach not only simplifies the design and manufacturing processes of locomotion robots, but also substantially improves their energy efficiency, which is comparable to those of animals. In this paper, we explain the novelty and principles of this approach through the four representative case studies that we have been exploring, and discuss challenges and perspectives toward the future. © 2012 IEEE.
Looking for publications? All our latest publications are listed here, and you can also use our search functions to help find what you are looking for. Power users might also want to consider searching on the EPFL Infoscience site which provides advanced publication search capabilities.
Most of the conventional legged robots are based on rigid body parts connected by high-torque actuators and a sophisticated control scheme to achieve stable running locomotion. The energy-efficiency of such robots is roughly 10-100 times lower than that of animals. Recently, there has been an increasing interest in designing compliant robots which exploit body dynamics for adaptive locomotion. It was shown that freevibration of elastic mechanical structures can generate energyefficient hopping/walking behavior. However, the velocity of such robots is very low. From this perspective, this paper presents a novel design strategy for running robots which makes use of torsional vibration of elastic beam. We propose a simple physical model which can be used to obtain the resonance frequency of the robot during stance phase. Moreover, this model can represent the dynamic behavior of such a robot during running locomotion.
There has been a long-standing debate on the question of how basic reflexive behaviours in mammals come about. Recently, it has been hypothesized that soft musculoskeletal interactions, such as intrinsic passive dynamics, might play a crucial role in the development of motor control at an early developmental stage. Inspired by the developmental processes, this paper explores a learning framework that enables us to systematically investigate the sensorimotor activity induced in soft musculoskeletal systems, as well as to self-organize a set of decentralized controllers analogue to spinal reflexes in mammals. This paper particularly focuses on three reflexes: the Myotatic reflex, the Reciprocal Inhibition reflex and the Reverse Myotatic reflex. We tested our framework in a simulated pair of soft muscles assembled in an agonist-antagonist arrangement. Our results show that the reflex circuitry as well as the reflex behaviour obtained are consistent with those observed in the mammal spinal cord.
Recent results in spinal research are challenging the historical view that the spinal reflexes are mostly hardwired and fixed behaviours. In previous work we have shown that three of the simplest spinal reflexes could be self-organised in an agonist-antagonist pair of muscles. The simplicity of these reflexes is given from the fact that they entail at most one interneuron mediating the connectivity between afferent inputs and efferent outputs. These reflexes are: the Myotatic, the Reciprocal Inibition and the Reverse Myotatic reflexes. In this paper we apply our framework to a simulated 2D leg model actuated by six muscles (mono- and bi-articular). Our results show that the framework is successful in learning most of the spinal reflex circuitry as well as the corresponding behaviour in the more complicated muscle arrangement.
There has been an increasing interest in the use of unconventional materials and morphologies in robotic systems because the underlying mechanical properties (such as body shapes, elasticity, viscosity, softness, density and stickiness) are crucial research topics for our in-depth understanding of embodied intelligence. The detailed investigations of physical system-environment interactions are particularly important for systematic development of technologies and theories of emergent adaptive behaviors. Based on the presentations and discussion in the Future Emerging Technology (fet11) conference, this article introduces the recent technological development in the field of soft robotics, and speculates about the implications and challenges in the robotics and embodied intelligence research. (C) Selection and peer-review under responsibility of FET11 conference organizers and published by Elsevier B.V.
A single leg hopping robot has been constructed which includes a clutch in series with the hip motor and a prototype Linear Multi-Modal Actuator (LMMA) at the knee. The single leg will be used to test how the different actuation methods can improve the behavioural diversity of the robot.
Future neuroprosthetic devices, in particular upper limb, will require decoding and executing not only the user’s intended movement type, but also when the user intends to execute the movement. This work investigates the potential use of brain signals recorded non-invasively for detecting the time before a self-paced reaching movement is initiated which could contribute to the design of practical upper limb neuroprosthetics. In particular, we show the detection of self-paced reaching movement intention in single trials using the readiness potential, an electroencephalography (EEG) slow cortical potential (SCP) computed in a narrow frequency range (0.1-1 Hz). Our experiments with 12 human volunteers, two of them stroke subjects, yield high detection rates prior to the movement onset and low detection rates during the non-movement intention period. With the proposed approach, movement intention was detected around 500 ms before actual onset, which clearly matches previous literature on readiness potentials. Interestingly, the result obtained with one of the stroke subjects is coherent with those achieved in healthy subjects, with single-trial performance of up to 92% for the paretic arm. These results suggest that, apart from contributing to our understanding of voluntary motor control for designing more advanced neuroprostheses, our work could also have a direct impact on advancing robot-assisted neurorehabilitation.
Programming by Demonstration offers an intu- itive framework for teaching robots how to perform various tasks without having to preprogram them. It also offers an intuitive way to provide corrections and refine teaching during task execution. Previously, mostly position constraints have been taken into account when teaching tasks from demonstrations. In this work, we tackle the problem of teaching tasks that require or can benefit from varying stiffness. This extension is not trivial, as the teacher needs to have a way of communicating to the robot what stiffness it should use. We propose a method by which the teacher can modulate the stiffness of the robot in any direction through physical interaction. The system is incremental and works online, so that the teacher can instantly feel how the robot learns from the interaction. We validate the proposed approach on two experiments on a 7-Dof Barrett WAM arm.
Neural signatures of humans’ movement intention can be exploited by future neuroprosthesis. We propose a method for detecting self-paced upper limb movement intention from brain signals acquired with both invasive and noninvasive methods. In the first study with scalp electroencephalograph (EEG) signals from healthy controls, we report single trial detection of movement intention using movement related potentials (MRPs) in a frequency range between 0.1 to 1 Hz. Movement intention can be detected above chance level (p<0.05) on average 460 ms before the movement onset with low detection rate during the on-movement intention period. Using intracranial EEG (iEEG) from one epileptic subject, we detect movement intention as early as 1500 ms before movement onset with accuracy above 90% using electrodes implanted in the bilateral supplementary motor area (SMA). The coherent results obtained with non-invasive and invasive method and its generalization capabilities across different days of recording, strengthened the theory that self-paced movement intention can be detected before movement initiation for the advancement in robot-assisted neurorehabilitation.