Central Pattern Generator Research Articles

Motion control of legged robots based on gradient central pattern generators

Motion control of legged robots based on gradient central pattern generators

Stable Walking of a Biped Robot Controlled by Central Pattern Generator Using Multivariate Linear Mapping.

In order to improve the walking stability of a biped robot in multiple scenarios and reduce the complexity of the Central Pattern Generator (CPG) model, a new CPG walking controller based on multivariate linear mapping was proposed. At first, in order to establish a dynamics model, the lower limb mechanical structure of the biped robot was designed. According to the Lagrange and angular momentum conservation method, the hybrid dynamic model of the biped robot was established. The initial value of the robot's passive walking was found by means of Poincaré mapping and cell mapping methods. Then, a multivariate linear mapping model was established to form a new lightweight CPG model based on a Hopf oscillator. According to the parameter distribution of the new CPG model, a preliminary parameter-tuning idea was proposed. At last, the joint simulation of MATLAB and V-REP shows that the biped robot based on the new CPG control has a stable periodic gait in flat and uphill scenes. The proposed method could improve the stability and versatility of bipedal walking in various environments and can provide general CPG generation and a tuning method reference for robotics scholars.

Enhancing postural stability in musculoskeletal quadrupedal locomotion through tension feedback for CPG-based controller.

A central pattern generator (CPG)-based controller enhances the adaptability of quadrupedal locomotion, for example, by controlling the trunk posture. The conventional CPG-based controllers with attitude control often utilized the posture angle as feedback information. However, if the robot's body is as soft as a musculoskeletal structure, it can measure the over-tilting of the trunk based on proprioceptive information of the muscles. In general, proprioceptive information such as muscle tension changes more rapidly than posture angle information. Therefore, a feedback loop based on proprioceptive information has great potential to respond to sudden disturbances that occur during locomotion over uneven terrain. In this research, we proposed a CPG-based controller utilizing the tension of soft pneumatic artificial muscles (PAMs). Musculoskeletal quadruped robots driven by PAMs are so soft, which prevents over-extension of the leg and over-tilting of the trunk to some extent. In addition, tension, one of the proprioceptive information of PAMs, exhibits high sensitivity to changes in trunk posture because the soft body's motion easily changes due to over-titing of the trunk. To validate the efficacy of the proposed controller, we conducted numerical simulations with a simple quadruped model and experiments with a musculoskeletal quadruped robot. As a result, the tension feedback was effective for posture stabilization when the robot's locomotion was subjected to disturbances. Moreover, the tension feedback was effective in improving the running velocity over uneven terrain. These results will enhance the locomotion capability of musculoskeletal quadruped robots, advancing their practical application.&#xD.

Bionic Walking Control of a Biped Robot Based on CPG Using an Improved Particle Swarm Algorithm

In the domain of bionic walking control for biped robots, optimizing the parameters of the central pattern generator (CPG) presents a formidable challenge due to its high-dimensional and nonlinear characteristics. The traditional particle swarm optimization (PSO) algorithm often converges to local optima, particularly when addressing CPG parameter optimization issues. To address these challenges, one improved particle swarm optimization algorithm aimed at enhancing the stability of the walking control of biped robots was proposed in this paper. The improved PSO algorithm incorporates a spiral function to generate better particles, alongside optimized inertia weight factors and learning factors. Evaluation results between the proposed algorithm and comparative PSO algorithms were provided, focusing on fitness, computational dimensions, convergence rates, and other metrics. The biped robot walking validation simulations, based on CPG control, were implemented through the integration of the V-REP (V4.1.0) and MATLAB (R2022b) platforms. Results demonstrate that compared with the traditional PSO algorithm and chaotic PSO algorithms, the performance of the proposed algorithm is improved by about 45% (two-dimensional model) and 54% (four-dimensional model), particularly excelling in high-dimensional computations. The novel algorithm exhibits a reduced complexity and improved optimization efficiency, thereby offering an effective strategy to enhance the walking stability of biped robots.

Color-induced cognitive conflicts affect muscle activity prior to gait initiation in the Go/No-go task.

In traffic rule, green/blue means go, and red means stop. It has been shown that this prior knowledge about traffic signal colors can affect reaction times (RTs). For example, RTs are longer when responding to a red "Go" signal and withholding the response to a blue "No-go" signal (Red Go/Blue No-go task) than when responding to a blue "Go" signal and withholding the response to a red "No-go" signal (Blue Go/Red No-go task), when responses are provided by button press. However, it remains unknown whether this holds in different actions. The aim of this study was to investigate the effect of prior knowledge of color on gait initiation in a Go/No-go task. Seventeen participants performed Green Go/Red No-go and Red Go/Green No-go tasks, in which they stepped forward from a force plate in response to a green or red signal and withhold the response to red or green signal, respectively. We recorded the center of pressure (COP) and electromyogram (EMG) from the bilateral tibialis anterior muscles during gait initiation. The onset of COP movement and toe-off time as well as COP displacements did not differ between the Go/No-go tasks. The EMG onset for the stance leg was delayed in the Red Go/Green No-go than Green Go/Red No-go task. These findings suggest that the conflict between prior knowledge of color related to traffic rule and the meaning of the stimulus color affects muscle activity but not COP characteristics during gait initiation, highlighting two distinct motor control mechanisms, where the initial phase is influenced by cognitive load while the subsequent phase remains unaffected. This dissociation suggests that the later phase of gait initiation relies on robust spinal loops and central pattern generators, which are less influenced by cognitive factors such as prior knowledge.

Personalized assist-as-needed dressing assistance robot without human modeling using Rowat-Selverston CPG controller

Dressing is a critical component of Activity of Daily Living (ADLs). The development of assistive technology for dressing tasks is urgently needed from the viewpoint of privacy, for example, to prevent the wearer from being seen in the nude. Despite their importance, these technologies are underutilized in garment donning compared to other ADLs. The reasons for the lack of utilization of assistive technology are that the task is complex, as it involves handling individuals with large individual differences in height and symptoms and requires manipulation of flexible clothing. Addressing these variances presents a formidable academic challenge. Our previous study identified periodic patterns in human movements during robot-assisted dressing and noted individual variability in these patterns. Capitalizing on this finding, we propose a novel control strategy for a robot that adapts to individual needs during dressing, employing the ‘Assist-As-Needed’ (AAN) principle from physical therapy. The proposed control method uses Central Pattern Generators (CPG) that can be synchronously controlled in response to external forces, enabling model-free control without human modeling. We utilize the Rowat-Selverston CPG model, which is recognized for its adaptive response to human motions in human-robot interactions such as with handshaking robots. A simulator of the Rowat-Selverston CPG was created. The output of the CPG was confirmed by inputting the data set obtained in previous studies, and the parameters were adjusted. The CPG output was then applied to the recorded target joint trajectory for the dressing assistance. We prepared a replay of the default trajectory, a trajectory with simple harmonic motion applied, and a trajectory with CPG output applied, and confirmed whether individual adaptation according to the AAN was possible through subject experiments. The experimental results showed that the Rowat-Selverston CPG control method enables individual adaptive dressing assistance according to the AAN principle. This study summarizes these methods and results and contributes to the realization of an individually adaptive system that follows the AAN principle, especially in developing assistive technology.

EFFECTS OF SPINAL TRANSECTION AND LOCOMOTOR SPEED ON MUSCLE SYNERGIES OF THE CAT HINDLIMB.

It was suggested that during locomotion, the nervous system controls movement by activating groups of muscles, or muscle synergies. Analysis of muscle synergies can reveal the organization of spinal locomotor networks and how it depends on the state of the nervous system, such as before and after spinal cord injury, and on different locomotor conditions, including a change in speed. The goal of this study was to investigate the effects of spinal transection and locomotor speed on hindlimb muscle synergies and their time-dependent activity patterns in adult cats. EMG activities of 15 hindlimb muscles were recorded in 9 adult cats of either sex during tied-belt treadmill locomotion at speeds of 0.4, 0.7, and 1.0 m/s before and after recovery from a low thoracic spinal transection. We determined EMG burst groups using cluster analysis of EMG burst onset and offset times and muscle synergies using non-negative matrix factorization. We found five major EMG burst groups and five muscle synergies in each of six experimental conditions (2 states × 3 speeds). In each case, the synergies accounted for at least 90% of muscle EMG variance. Both spinal transection and locomotion speed modified subgroups of EMG burst groups and the composition and activation patterns of selected synergies. However, these changes did not modify the general organization of muscle synergies. Based on the obtained results, we propose an organization for a pattern formation network of a two-level central pattern generator that can be tested in neuromechanical simulations of spinal circuits controlling cat locomotion.

Multistability of bursting rhythms in a half-center oscillator and the protective effects of synaptic inhibition.

For animals to meet environmental challenges, the activity patterns of specialized oscillatory neural circuits, central pattern generators (CPGs), controlling rhythmic movements like breathing and locomotion, are adjusted by neuromodulation. As a representative example, the leech heartbeat is controlled by a CPG driven by two pairs of mutually inhibitory interneurons, heart interneuron (HN) half-center oscillators (HCO). Experiments and modeling indicate that neuromodulation of HCO navigates this CPG between dysfunctional regimes by employing a co-regulating inverted relation; reducing Na+/K+ pump current and increasing hyperpolarization-activated (h-) current. Simply reducing pump activity or increasing h-current leads to either seizure-like bursting or an asymmetric bursting dysfunctional regime, respectively. Here, we demonstrate through modeling that, alongside this coregulation path, a new bursting regime emerges. Both regimes fulfill the criteria for functional bursting activity. Although the cycle periods and burst durations of these patterns are roughly the same, the new one exhibits an intra-burst spike frequency that is twice as high as the other. This finding suggests that neuromodulation could introduce additional functional regimes with higher spike frequency, and thus more effective synaptic transmission to motor neurons. We found that this new regime co-exists with the original bursting. The HCO can be switched between them by a short pulse of excitatory or inhibitory conductance. In this domain of coexisting functional patterns, an isolated cell model exhibits only one regime, a severely dysfunctional plateau-containing, seizure-like activity. This aligns with widely reported notion that deficiency of inhibition can cause seizures and other dysfunctional neural activities. We show that along the coregulation path of neuromodulation, the high excitability of the single HNs induced by myomodulin is harnessed by mutually inhibitory synaptic interactions of the HCO into the functional bursting pattern.

Walking on different inclines affects gait symmetry differently in the anterior-posterior and vertical directions: implication for future sensorimotor training

A symmetric gait pattern in humans reflects near-identical movement in bilateral limbs during walking. However, little is known about how gait symmetry changes on different inclines. This study aimed to address this knowledge gap using the central pattern generator and internal model hypotheses. Eighteen healthy young adults underwent five 2-minute walking trials (inclines of +15%, +8%, 0%, −8%, and −15%). Dependent variables included step time, step length, step width, maximum heel clearance, time to peaks of maximum heel clearance, their corresponding coefficients of variation (CV), and respective symmetry indices (SI). Significant differences were observed in SI of step length (p = .022), step length variability (p < .001), step width variability (p =.001), maximum heel clearance (p < .001), and maximum heel clearance variability (p = .049). Compared to level walking, walking at −8% and −15% inclines increased SI of step length (p = .011, p = .039 respectively) but decreased SI of maximum heel clearance (p = .025, p = .019 respectively). These observations suggested that incline walking affected gait symmetry differently, possibly due to varied internal models used in locomotion. Downhill walking improved vertical gait symmetry but reduced anterior-posterior symmetry compared to level walking. Downhill walking may be a preferable rehabilitation protocol for enhancing gait symmetry, as it activates internal model controls. Even slight downhill inclines could increase active control loading, beneficial for the elderly and those with impaired gait.

Finite-Time Line-of-Sight Guidance-Based Path-Following Control for a Wire-Driven Robot Fish.

This paper presents an adaptive line-of-sight (LOS) guidance method, incorporating a finite-time sideslip angle observer to achieve precise planar path tracking of a bionic robotic fish driven by LOS. First, an adaptive LOS guidance method based on real-time cross-track error is presented. To mitigate the adverse effects of the sideslip angle on tracking performance, a finite-time observer (FTO) based on finite-time convergence theory is employed to observe the time-varying sideslip angle and correct the target yaw. Subsequently, classical proportional-integral-derivative (PID) controllers are utilized to achieve yaw tracking, followed by static and dynamic yaw angle experiments for evaluation. Finally, the yaw-tracking-based path-tracking control strategy is applied to the robotic fish, whose motion is generated by an improved central pattern generator (CPG) and equipped with a six-axis inertial measurement unit for real-time swimming direction. Quantitative comparisons in tank experiments validate the effectiveness of the proposed method.

Bifurcation analysis of a two-neuron central pattern generator model for both oscillatory and convergent neuronal activities.

The neural oscillator model proposed by Matsuoka is a piecewise affine system that exhibits distinctive periodic solutions. Although such typical oscillation patterns have been widely studied, little is understood about the dynamics of convergence to certain fixed points and bifurcations between the periodic orbits and fixed points in this model. We performed fixed point analysis on a two-neuron version of the Matsuoka oscillator model, the result of which explains the mechanism of oscillation and the discontinuity-induced bifurcations such as subcritical/supercritical Hopf-like, homoclinic-like and grazing bifurcations. Furthermore, it provided theoretical predictions concerning a logarithmic oscillation-period scaling law and noise-induced oscillations observed around those bifurcations. These results are expected to underpin further investigations into oscillatory and transient neuronal activities concerning central pattern generators.

Role of the pontine respiratory group in the suppression of cough by codeine in cats

Codeine was microinjected into the area of the Kölliker-Fuse nucleus and the adjacent lateral parabrachial nucleus, within the pontine respiratory group in 8 anesthetized cats. Electromyograms (EMGs) of the diaphragm (DIA) and abdominal muscles (ABD), esophageal pressures (EP), and blood pressure were recorded and analyzed during mechanically induced tracheobronchial cough. Unilateral microinjections of 3.3 mM codeine (3 injections, each 37 ± 1.2 nl) had no significant effect on the cough number. However, the amplitudes of the cough ABD EMG, expiratory EP and, to a lesser extent, DIA EMG were significantly reduced. There were no significant changes in the temporal parameters of the cough. Control microinjections of artificial cerebrospinal fluid in 6 cats did not show a significant effect on cough data compared to those after codeine microinjections. Codeine-sensitive neurons in the rostral dorsolateral pons contribute to controlling cough motor output, likely through the central pattern generator of cough.

Sensorimotor control of robots mediated by electrophysiological measurements of fungal mycelia.

Living tissues are still far from being used as practical components in biohybrid robots because of limitations in life span, sensitivity to environmental factors, and stringent culture procedures. Here, we introduce fungal mycelia as an easy-to-use and robust living component in biohybrid robots. We constructed two biohybrid robots that use the electrophysiological activity of living mycelia to control their artificial actuators. The mycelia sense their environment and issue action potential-like spiking voltages as control signals to the motors and valves of the robots that we designed and built. The paper highlights two key innovations: first, a vibration- and electromagnetic interference-shielded mycelium electrical interface that allows for stable, long-term electrophysiological bioelectric recordings during untethered, mobile operation; second, a control architecture for robots inspired by neural central pattern generators, incorporating rhythmic patterns of positive and negative spikes from the living mycelia. We used these signals to control a walking soft robot as well as a wheeled hard one. We also demonstrated the use of mycelia to respond to environmental cues by using ultraviolet light stimulation to augment the robots' gaits.

Inference technique for the synaptic conductances in rhythmically active networks and application to respiratory central pattern generation circuits.

Unraveling synaptic interactions between excitatory and inhibitory interneurons within rhythmic neural circuits, such as central pattern generation (CPG) circuits for rhythmic motor behaviors, is critical for deciphering circuit interactions and functional architecture, which is a major problem for understanding how neural circuits operate. Here we present a general method for extracting and separating patterns of inhibitory and excitatory synaptic conductances at high temporal resolution from single neuronal intracellular recordings in rhythmically active networks. These post-synaptic conductances reflect the combined synaptic inputs from the key interacting neuronal populations and can reveal the functional connectome of the active circuits. To illustrate the applicability of our analytic technique, we employ our method to infer the synaptic conductance profiles in identified rhythmically active interneurons within key microcircuits of the mammalian (mature rat) brainstem respiratory CPG and provide a perspective on how our approach can resolve the functional interactions and circuit organization of these interneuron populations. We demonstrate the versatility of our approach, which can be applied to any other rhythmic circuits where conditions allow for neuronal intracellular recordings.

Functional involvement of the sternohyoid muscle during breathing and swallowing in rats.

The sternohyoid muscle depresses the hyoid bone, but it is unclear whether the muscle contributes to respiratory and swallowing mechanisms. This study aimed to clarify whether the sternohyoid muscle participates in the respiration and swallowing reflex and how the activity is modulated in two conditions: with airway stenosis and with a fixed sternohyoid muscle length. Electromyographic activity in the sternohyoid, digastric, thyrohyoid, and diaphragm muscles was recorded in anesthetized rats. The sternohyoid muscle activity was observed in the inspiratory phase and during swallowing, and was well coordinated with digastric and thyrohyoid muscle activity. With airway stenosis, the respiratory activity per respiratory cycle was facilitated in all assessed muscles but the facilitation of activity per second occurred only in the digastric, thyrohyoid, and sternohyoid muscles. With airway stenosis, the swallowing activity was facilitated only in the digastric muscle but not in the thyrohyoid and sternohyoid muscles. Swallowing activity was not observed in the sternohyoid muscle in the condition with the sternohyoid muscle length fixed, although increased inspiratory activity remained. The current results suggest that 1) the sternohyoid muscle is slightly activated in the inspiratory phase, 2) the effect of airway stenosis on respiratory function may differ between the upper airway muscles and diaphragm, and 3) swallowing activity in the sternohyoid muscle is not dominantly controlled by the swallowing central pattern generator but instead occurs as a myotatic reflex.NEW & NOTEWORTHY We found that the sternohyoid muscle was activated in the inspiratory phase. However, increased airway resistance had different effects on the extrathoracic muscles than on the diaphragm. The swallowing activity of the sternohyoid disappeared when the muscle length was fixed. These findings suggest that the sternohyoid muscle may be activated not by the swallowing central pattern generator but as a myotatic reflex.

Optimization of swimming mode for elongated undulating fin using multi-agent deep deterministic policy gradient

Optimizing speed and propulsive efficiency are the most crucial survival skills for biomimetic robots. This paper investigates a swimming mode controller inspired by the black Knifefish to govern the fast-swimming gait with high propulsive efficiency for an elongated undulating fin robot. The proposed swimming mode controller is composed of a couple of Hopf oscillator-based central pattern generators (CPG) to generate the moving gait of robotic fish and a novel variant of Reinforcement Learning (RL) known as Multi-Agent Deep Deterministic Policy Gradient (MA-DDPG) for optimizing the propulsive efficiency. The proposed swimming controller facilitates the autonomous optimization of the oscillatory amplitude of the robotic fish to improve its propulsive efficiency. The proposed MA-DDPG demonstrates an aptitude for functioning within mixed cooperative-competitive environments. Furthermore, it effectively mitigates the drawback of zero amplitude in the updating process of conventional reinforcement learning (RL) methodologies. These findings highlight the potential utility of the MA-DDPG in optimizing the performance of multi-agent systems in dynamic, real-world scenarios. The simulation results show that the undulating fin robot reaches a maximum thrust of 0.9 N with a propulsive efficiency of 12.48 %, which is higher than that of traditional reinforcement learning methods.

A model-free method to learn multiple skills in parallel on modular robots

Legged robots are well-suited for deployment in unstructured environments but require a unique control scheme specific for their design. As controllers optimised in simulation do not transfer well to the real world (the infamous sim-to-real gap), methods enabling quick learning in the real world, without any assumptions on the specific robot model and its dynamics, are necessary. In this paper, we present a generic method based on Central Pattern Generators, that enables the acquisition of basic locomotion skills in parallel, through very few trials. The novelty of our approach, underpinned by a mathematical analysis of the controller model, is to search for good initial states, instead of optimising connection weights. Empirical validation in six different robot morphologies demonstrates that our method enables robots to learn primary locomotion skills in less than 15 minutes in the real world. In the end, we showcase our skills in a targeted locomotion experiment.

Intelligent Control Strategy for Robotic Manta via CPG and Deep Reinforcement Learning

The robotic manta has attracted significant interest for its exceptional maneuverability, swimming efficiency, and stealthiness. However, achieving efficient autonomous swimming in complex underwater environments presents a significant challenge. To address this issue, this study integrates Deep Deterministic Policy Gradient (DDPG) with Central Pattern Generators (CPGs) and proposes a CPG-based DDPG control strategy. First, we designed a CPG control strategy that can more precisely mimic the swimming behavior of the manta. Then, we implemented the DDPG algorithm as a high-level controller that adaptively modifies the CPG’s control parameters based on the real-time state information of the robotic manta. This adjustment allows for the regulation of swimming modes to fulfill specific tasks. The proposed strategy underwent initial training and testing in a simulated environment before deployment on a robotic manta prototype for field trials. Both further simulation and experimental results validate the effectiveness and practicality of the proposed control strategy.

Sulfonylurea receptor coupled conductances alter the performace of two central pattern generating circuits in Cancer borealis.

Neuronal activity and energy supply must maintain a fine balance for neuronal fitness. Various channels of communication between the two could impact network output in different ways. Sulfonylurea receptors (SURs) are a modification of ATP-binding cassette proteins (ABCs) that confer ATP-dependent gating on their associated ion channels. They are widely expressed and link metabolic states directly to neuronal activity. The role they play varies in different circuits, both enabling bursting and inhibiting activity in pathological conditions. The crab, Cancer borealis, has central patterns generators (CPGs) that fire in rhythmic bursts nearly constantly and it is unknown how energy availability influences these networks. The pyloric network of the stomatogastric ganglion (STG) and cardiac ganglion (GC) control rhythmic contractions of the foregut and heart respectively. Pharmacological manipulation of SURs results in opposite effects in the two CPGs. Neuronal firing completely stops in the STG when SUR-associated channels are open, and firing increases when the channels are closed. This results from a decrease in the excitability of pyloric dilator (PD) neurons, which are a part of the pacemaker kernel. The neurons of the CG, paradoxically, increase firing within bursts when SUR-associated channels are opened, and bursting slows when SUR-associated channels are closed. The channel permeability and sensitivities analyses present novel SUR-conductance biophysics, which nevertheless change activity in ways reminiscent of the predominantly studied mammalian receptor/channels. We suggest that SUR-associated conductances allow different neurons to respond to energy states in different ways through a common mechanism.

SYNLOCO‐VE: Synthesizing central pattern generator with reinforcement learning and velocity estimator for quadruped locomotion

AbstractIt is a challenging task to learn a robust and natural locomotion controller for quadruped robots at different terrains and velocities. In particular, the locomotion learning task will be even more difficult for the case with no exteroceptive sensors. In this article, the learning‐based locomotion control is, therefore, investigated for quadruped robots only using proprioceptive sensors. A new framework called SYNLOCO‐VE is proposed by synthesizing a feedforward gait planner, a trunk velocity estimator, and reinforcement learning (RL). The feedforward gait planner is developed based on the well‐known central pattern generator, but it can change the foot length for improved velocity tracking performance. The trunk velocity estimator is designed based on deep learning, which estimates the trunk velocity using historical data from proprioceptive sensors. The introduction of the trunk velocity estimator can mitigate the influence of the partial observation issue due to the lack of exteroceptive sensors. RL is employed to learn a feedback controller to regulate the robot gaits using feedback from proprioceptive sensors and the trunk velocity estimation. In the proposed framework, the feedforward gait planner can also guide the training process of RL, thus resulting in more stable and faster policy learning. Ablation studies are provided to demonstrate the efficiency of different modules in the proposed design. Extensive experiments are performed using a quadruped robot Go1, which only has proprioceptive sensors. The proposed framework is able to learn robust and stable locomotion at different terrains and tasks. Experimental comparisons are also conducted to illustrate the advantages of the proposed design over the state‐of‐the‐art methods.