User:Jmaeng/sandbox

Locomotion robophysics is a term defined in to establish a discipline that studies the physics of moving systems. It is an interdisciplinary scientific field testing hypotheses and exploring physical principles of locomotion using robots as simplified model locomotors. This can include the development of systematically controlled substrates, template-anchor systems, and experiments investigating the parameter space for non-failing locomotion. Once the principles are developed in the robophysical system, they can aid in the discovery of new concepts in physics, biology, and control engineering. For instance, dynamics that arise from robot-granular substrate interactions can lead to advances in soft matter physics.

Basic idea
Robophysics has its roots in the template-anchor viewpoint, which systematically reduces the dimension of locomoting systems by abstraction. This viewpoint is founded on the success of the spring-loaded inverted pendulum (SLIP) model, which has been instrumental in advancing our understanding of legged locomotion on the ground. By combining the SLIP template with neuromechanical hypotheses, we have gained important insights into the locomotion of animals and robots. This has led to the idea of applying the same approach to other complex systems.

The robophysical approach conjugates templates and anchors with systematic experiments. By exploring the parameter space, robophysics study both successful and unsuccessful locomotion strategies. For instance, the experiment varying the actuator frequency and phase of a vertical jumping robot with actuated mass-spring arrangement revealed two distinct jumping modes emerge as a result of nonresonant transient dynamics.

As such, robotic experiments take an important role because they are simpler than real living organisms while serving as physical simulations where we could obtain locomotor behaviors without knowing all the physical laws behind them.

Examples
The implementation of the robophysical approach is a challenging task that demands a considerable amount of effort and creativity. One needs to simplify the system, transform it into a robot, and derive meaningful data from the intricate outcomes of its components. Thus, studying successful examples of this approach can be beneficial for better comprehension.

Study embodied intelligence
Embodied intelligence is a theory that stresses the beauty of cheap design replacing complex perception and control. One example is passive collision avoidance with low-stiffness limbs in small insects and running birds. Instead of sensing the contact and replanning the swing leg trajectory accordingly, many legged ambulators change their limb stiffness and eventually response to collisions.

studied the role of limb stiffness, an anchor that legged-ambulators use, by systematically varying the limb stiffness of a robot leg. Here, one robot leg was used such that effects other than leg compliance, such as a change in body center of mass or leg coordination, could be neglected. Also, they tested two different control methods on the same robot leg, which could not be done on real animals or insects. As a result, they discover the importance of anisotropic compliance when a leg needs to negotiate its step due to obstacles. In short, a systematic study of the parameter space not only reveals the existence of a certain anchor but also how it enhances performance.

Advancement in soft matter physics
In numerous instances, the physical laws governing the environment in which the self-deforming ambulator moves are as intricate as the ambulator itself. This is due to the fact that the majority of the substances in the environment, such as sand, water, and wind, possess non-linear attributes.

created a systematic testing bed for heterogeneous granular media and the discrete element simulation tool based on experimental validation. The research team applied Resistive Force Theory (RFT), that is originally developed for viscous fluid and discovered that RFT could well predict the interaction between leg and granular media in the quasistatic locomotor regime. In brief, the robophysical approach deepens the understanding of terrestrial substrates with unknown equations of motion.

Gait analysis
Locomotion encompasses the coordination of muscles and limbs that could either be achieved by global coupling by brain commands or local reflexes of each limb.

One way of studying the neuromechanics of locomotion is to reconstruct the robot and its movement, then study the robot's parameter space. studied centralized locomotion control of extinct species, Orobates, using fossil records, a bio-inspired robot, and kinematic/dynamic simulations. They explored biomechanically possible gaits, morphologies, and footprints of Orobates and tested their methods with existing descendants of the species.

The authors abstracted the Orobates gaits to a Sprawling Gait Space (SGS) consisting of body height, lateral bending of the vertebral column, and long-axis rotation and retraction in proximal limb joints, and used power expenditure, balance, footstep precision, and ground reaction force (GRF) as gait evaluation metrics. These selections could be explored further, both as templates and anchors, depending on the behavior of interest.

Challenges
Possible challenges include :
 * 1) Relevance: relevance of the hypotheses to biology. (Are the assumptions biologically plausible?)
 * 2) Level: the scale of the model from atoms to societies. (We can't understand the lift generation of birds by only zooming into the feather structure.)
 * 3) Generality: the range of biological systems that the model can depict. (Theories for bacteria swimming at low Reynolds regime cannot be applied to whale swimming.)
 * 4) Abstraction: level of detail incorporated into the model compared to the target. (e.g., inverted pendulum, SLIP, SLIP with Swing Legs (SLIP-SL), SLIP with variable leg stiffness (V-SLIP), etc.)
 * 5) Structural accuracy: the extent to which the model represents the underlying mechanisms of the behavior. (how can we overcome the discrepancies between artificial and natural motors and sensors?)
 * 6) Performance match: the extent to which the model's behavior matches the target behavior (Need performance evaluation metrics as well.)
 * 7) Medium: different forms a model might take (Including models with the same materials as its target, models that share some physical properties with their targets, or models in computer programs.)