Opening Talk | Monday, 11th

Abstract: We have developed a technology that can decompose the surface EMG signal and identify the shape and firing instances of individual motor unit action potentials that contribute to the surface EMG signal with an average accuracy of 95%. Using this technology we have document a hierarchal construct of the motor-unit firing rate, whereby the firing rate of any motor unit is inversely proportional to the force at which it is recruitment. Thus, at any time and force, the firing rate of subsequently recruited greater force-twitch faster-fatiguing motor units is less than the previously recruited. This construct opposes the previously belief held for over 50 years.

Our findings have the following implications: 1) the control of motor units within a muscle is not designed to maximize the force output of a muscle, but instead it seems to have optimized some combination of force magnitude and time duration. (The control scheme is well suited for the flight-and-fight response by providing the capacity to generate force and the capacity to sustain it.) 2) As the force level increase, the variability increases. 3) It provides a greater economy of force generation during ordinary functional daily activities, such as normal walking, that require bursts of low force levels generated over short periods of time. Such activities rely on the activation of lower threshold motor units that fire faster and tetanize relatively quickly to produce force at lower levels.

When we investigated the co-activation of synergist and antagonistic muscles around a joint, we found a substantial amount of cross-correlation between the firing rates of the two muscles. This finding indicates that muscles are activated as a functional group even at the motor unit level.

The implications of these findings for designing control systems for robots that intendend to mimic human based movement will be pointed out.

Keynote Talk

Abstract: One approach to the design of new devices and materials is to study how problems have been solved in nature and to adapt these solutions for our own uses. This “biomimetic” approach is currently being used at Tufts to develop a new class of robots fabricated from soft materials. These soft robots will perform tasks outside the capability of current robots including climbing textured surfaces, crawling along ropes and wires and burrowing into winding, confined spaces.

Making these machines move accurately will require the application of new concepts based on neuromechanics and embodiment (morphological computation). These conceptual breakthroughs also allow for the future production of machines that are entirely biosynthetic and biodegradable and that can be grown rather than assembled.

This talk will focus on recent discoveries in the locomotion of soft animals and the application of these findings to the development of moving machines that are highly deformable. These are the early prototypes of a new type of engineering based on controlling structures built entirely of soft materials.

Abstract: Robot designers are increasingly searching for ideas from biology. The talk will introduce such bio-inspired robots that embody the hypothesized principles from the insights obtained by animal studies. Through these examples, the intricate processes of design principle extraction will be discussed. Current research in the MIT biomimetics lab is centered on the development of a cheetah-inspired running robot. Three major associated research thrusts are optimum actuator design, biotensegrity structure design, and the impluse-based control architecture for stable galloping control. Each research component is guided by biomechanics of runners such as dogs and cheetahs capable of the fast traverse on rough and unstructured terrains.

Abstract: The emergence of social adaptability must be common interest between biologists and robotics engineers. In animals, social interaction is important factor for the decision making of behaviors. In order to understand how animals alter their behaviors on the demand of changing social environment, I have focus on aggressive behavior in insects. It is widely observed in animals that dominant hierarchy is established by agonistic behavior, through the complex interaction among physiological, motivational, and behavioral systems. Crickets exhibit intensive aggressive behavior when one encounters another male. The battle starts out slowly and escalates into a fierce struggle to establish dominant-subordinate relationship. We have investigated the neuronal mechanism underlying cricket aggressive behavior. Pharmacological experiments suggest that nitric oxide signaling mediates octopaminergic system in the brain, which in turn mediates aggressive motivation. Based on the results of experiments, we established dynamic behavior models and neurophysiological models to understand the mechanisms of social adaptability. We hypothesize that important mechanism underlying behavior adaptability is a multiple feedback structure that is composed of feedback loop in the nervous systems and through the social environment.

Tandem Talk

Abstract: The movements of animals exhibit great complexity. One reason for this complexity is that animals actively move and control many degrees of freedom of motion simultaneously. Another reason is that the prevalent features of the neural control system may change over time, for example when adapting to a certain "task" or context. In our tandem talk, we will address problems associated with experimental analysis and quantitative descriptions of natural, unrestrained locomotion. We will do this from two complementary perspectives, that of a behavioral neuroscientist and that of a control engineer. We will motivate the problem using simple examples of the neurobiology of multi-legged locomotion: the coordination of joints, body segments and limbs. After illustrating selected approaches of their analysis in neurobiology (Volker Dürr), we will contrast these approaches with corresponding analytical tools from systems engineering (Noah Cowan). Following this methodological preface we will present selected special cases of natural motion analysis in cockroaches and fishes, humans and stick insects.

In the “reverse engineering” component of our talk (N. Cowan), we will illustrate the power of simple control theoretic tools for the analysis of animal locomotion. In particular, we will examine the interplay between “top-down” input–output models of behavior with first-principles models of locomotor mechanics in discovering neural control algorithms.

In a final neurobiology section (V. Dürr), we will emphasize the role of natural variability of movement, and on how to identify invariances that are likely to reflect the properties of the underlying control mechanisms. Examples will comprise natural catching movements of humans and insect locomotion on variable terrain. For the latter, we will discuss findings based on large kinematic data sets allowing identification of invariant patterns in the natural statistics of locomotion across species. Using the example of bimodal step length distributions, we will discuss whether the two modes of stepping may indicate two mechanisms of step generation. This will be complemented by experimental findings on sensory monitoring of the interaction between body and substrate.

Abstract: Locomotor control is not limited to the commands of the central nervous system. The physical interactions between the body and the external environment play a crucial role in generating adaptive motion in biological and robotic systems. These interactions are often mediated by the passive mechanical properties of a moving body and define a set of intrinsic control mechanisms.

In the field of Biology, passive mechanical properties have been shown to be the first line of defense against destabilizing perturbations allowing biological systems to self-correct without nervous input. In addition, variation in the mechanical properties of muscles and tendons can determine the boundaries of locomotor performance in human and animal systems. Based on these biological findings, Pfeifer et al. have coined the term “morphological computation” to describe how the physical properties of the body provide a passive control mechanism in moving robots. Many researchers have demonstrated the potential of morphological computation for the generation of stable and versatile behavior in robotic systems.

In this tandem talk, we will discuss how passive properties in biological and robotic systems determine posture, gait and stability during locomotion.

Abstract: There are roughly two approaches for understanding bipedal locomotion: starting from observation of bipedal animals and from constructing and controlling of bipedal robots. In this tandem talk, Dr. Daley will discuss inferences about control of bipedal locomotion from observations of birds' behavior. She will introduce a simple bio-inspired leg control policy that has emerged from these observations. Prof. Hosoda will then talk about constructive approach toward understanding bipedal locomotion. He will introduce several muscular-skeletal biped robots imitating biological systems. Finally, they will conclude the talk discussing on simplification of the controller based on appropriate morphological design, and on understanding of the adaptability of the bipedal locomotion.

Public Lecture | Wednesday 13th March

Abstract: Animals provide remarkable models for elegant movement through extremely complex terrains. Robotics seminars often begin with envious video of mountain goats jumping around hill sides and insects moving seamlessly around numerous barriers to a goal. AMAM is unique in its serious attempt to bring together biologists who study these behaviors and roboticists who are trying to capture this kind of movement in their devices. The benefits of robots that could move with animal-like behavior are readily apparent. However, much of the work that is done in the field breaks the problem into smaller units. We see fabulous work on legged movement generating algorithms for stability and efficient propulsion but ignoring the role of higher centers. At the other extreme wheeled vehicles are connected to sensors that allow navigation around barriers but without the benefits of legs. Biologists likewise focuses upon small subsets of the problem. Neurobiologists have made great progress examining local control in spinal cords or thoracic ganglia by eliminating or ignoring descending brain control and restricting movement to a small subset of actions on a treadmill or a narrow track or studying rapidly moving behaviors that may occur with limited sensory intervention. Such research is clearly important and has been very successful in making progress in both fields. But at some point, we must consider the interactions between higher centers and local control that is necessary for the elegant movements that roboticists seek to capture. The issue is similar to progress made by an orchestra working on complex musical pieces. Individual sections must work on their own to reach proficiency on their passages. But the symphony only reaches its full potential when all the parts are brought together with all the complex interactions that make it wonderful.

In this talk, I will discuss the differences in behavior that occur in insect locomotion when brain circuits are present or absent and then describe the research that our laboratory is conducting to understand how these brain circuits function and interact with local control systems that include the reflexes and pattern generators of the thoracic ganglia to produce the kinds of behavior seen in those initial videos. The ultimate goal is both to understand the neural properties that underlie complex behavior and to generate the kinds of robotic control that would capture those behaviors with minimal external control from a driver.

• View
• Edit
• History
• Print
• Backlinks