Publications

2020

1. Robust and reusable self-organized locomotion of legged robots under adaptive physical and neural communications (Under review) T. Sun, Z, Dai, and P. Manoonpong Frontiers in Neural Circuits 2020 Abstract   Journal paper   Under Review

   Animals, such as cattle, can perform automatic sensorimotor coordination for versatile and elegant behaviors. Their self-organized movement exhibits an impression of adaptability, robustness, and memorability. In the past few decades, roboticists have tried to imitate such natural properties with artificial legged locomotion systems by using different approaches, including classic engineering control techniques, machine learning algorithms, and bio-inspired control mechanisms. However, the adaptive mechanisms underlying the natural properties of animals have not been completely realized in artificial legged systems so far. Based on the biological findings, the adaptive mechanisms seem to depend, to a great extent, not only on sensory feedback modulation (e.g., Tegotae-based control), which provides physical communication for distributed neural control circuits, but also on neural communication among the neural circuits and their combinations. From this point of view, we propose adaptive neural control to accomplish adaptive physical and neural communications. This control algorithm consists of distributed local central pattern generator (CPG)-based neural circuits for generating basic leg movements, an adaptive sensory feedback mechanism for generating self-organized phase relationships among the local CPG circuits, and an adaptive neural coupling mechanism for transferring and storing the formed phase relationships (a gait pattern) into the neural structure. This adaptive neural control was evaluated on a quadruped robot. As a result, it enables the robot to 1) rapidly and automatically form its gait (i.e., self-organized locomotion) within a few seconds, 2) memorize the gait for later recovery, and 3) robustly walk even when facing a sensory feedback malfunction. It also allows for changing the walking speed and direction of the robot for maneuverability. Moreover, the implementation of the adaptive physical and neural communications provides an option for understanding the mechanism of motor memory formation. As a whole, the study also demonstrates that the integration of the two world communications through the adaptive neural control is a powerful way to achieve robust and reusable self-organized locomotion in legged robots.

2. Distributed-force-feedback-based-reflex with online learning for adaptive quadruped motor control Sun Tao, Zhendong Dai, and Poramate Manoonpong. Neural Network 2020 Abstract   Journal paper [HTML]

   Biological motor control mechanisms, such as central pattern generators (CPGs), sensory feedback, reflexes, and motor learning, play a crucial role for adaptive locomotion in animals as well as legged robots. However, the interaction and integration among these mechanisms for generating efficient adaptive locomotion to deal with diverse terrains have not been fully realized. One issue is to achieve adaptive motor control for fast postural adaptation on various terrains. To address this issue, taking all these biological mechanisms into account, this study proposes a novel distributed force feedback-based reflex with online learning (DFRL). It integrates force sensory feedback, reflex, and learning to cooperate CPGs for producing adaptive motor commands, thereby leading to fast body posture adaptation for stable trotting of quadruped robots on various terrains. The DFRL is based on a simple neural network with plastic synapses modulated online by a fast dual integral learner (DIL). Compared to the traditional vestibular reflex, the proposed DFRL shows faster adaptation on a wide range of slopes and extensive generalization on different platforms. This is due to its neural plasticity. Experimental results on quadruped robots show that the DFRL can 1) automatically and quickly adapt the CPG patterns (motor commands) to the robots to perform appropriate body posture during locomotion; and 2) enable the robots to effectively accommodate various slope terrains including steep ones (i.e., \ang{\pm 35} for a small-sized robot with 0.2 m height and \ang{50} and \ang{-45} for a larger robot with 0.6 m height). Consequently, the robots driven by the DFRL can successfully perform efficient adaptive locomotion to deal with a complex terrain with diverse slopes.

3. A comparative study of adaptive interlimb coordination mechanisms for self-organized robot locomotion Sun Tao, Xiaofeng Xiong, Zhendong Dai, Dai Owaki, and Poramate Manoonpong. Frontiers in Robotics and AI 2020 Abstract [HTML] [PDF]   Journal paper

   Walking animals demonstrate impressive self-organized locomotion and adaptation to body property changes by skillfully manipulating their complicated and redundant musculoskeletal systems. Adaptive interlimb coordination plays a crucial role in this achievement. It has been identified that interlimb coordination is generated through dynamical interactions between the neural system, musculoskeletal system, and environment. Based on this principle, two classical interlimb coordination mechanisms (continuous phase modulation and phase resetting) have been proposed independently. These mechanisms use decoupled central pattern generators (CPGs) with sensory feedback, such as ground reaction forces (GRFs), to generate robot locomotion autonomously without predefining it (i.e., self-organized locomotion). A comparative study was conducted on the two mechanisms under decoupled CPG-based control implemented on a quadruped robot in simulation. Their characteristics were compared by observing their CPG phase convergence processes at different control parameter values. Additionally, the mechanisms were investigated when the robot faced various unexpected situations, such as noisy feedback, leg motor damage, and carrying a load. The comparative study reveals that the phase modulation and resetting mechanisms demonstrate satisfactory performance when they are subjected to symmetric and asymmetric GRF distributions, respectively. This work also suggests a strategy for the appropriate selection of adaptive interlimb coordination mechanisms under different conditions and for the optimal setting of their control parameter values to enhance their control performance.

4. Small-Sized Reconfigurable Quadruped Robot With Multiple Sensory Feedback for Studying Adaptive and Versatile Behaviors Sun Tao, Xiaofeng Xiong, Zhendong Dai, and Poramate Manoonpong. Frontiers in Neurorobotics 2020 Abstract [HTML] [PDF]   Journal paper

   Self-organization of locomotion characterizes the feature of automatically spontaneous gait generation without preprogrammed limb movement coordination. To study this feature in quadruped locomotion, we propose here a new open-source, small-sized reconfigurable quadruped robot, called Lilibot, with multiple sensory feedback and its physical simulation. Lilibot was designed as a friendly quadrupedal platform with unique characteristics, including light weight, easy handling, modular components, and multiple real-time sensory feedback. Its modular components can be flexibly reconfigured to obtain features such as different leg orientations for testing the effectiveness and generalization of self-organized locomotion control. Its multiple sensory feedback (i.e., joint angles, joint velocities, joint currents, joint voltages, and body inclination) can support vestibular reflexes and compliant control mechanisms for body posture stabilization and compliant behavior, respectively. To evaluate the performance of Lilibot, we implemented our developed adaptive neural controller on it. The experimental results demonstrated that Lilibot can autonomously and rapidly exhibit adaptive and versatile behaviors, including spontaneous self-organized locomotion (i.e., adaptive locomotion) under different leg orientations, body posture stabilization on a tiltable plane, and leg compliance for unexpected external load compensation. To this end, we successfully developed an open-source, friendly, small-sized, and lightweight quadruped robot with reconfigurable legs and multiple sensory feedback that can serve as a generic quadrupedal platform for research and education in the fields of locomotion, vestibular reflex-based, and compliant control.

2018

1. Adaptive neural control for self-organized locomotion and obstacle negotiation of quadruped robots Tao Sun, Donghao Shao, Zhendong Dai, Poramate Manoonpong 27th IEEE International Symposium on Robot and Human Interactive Communication (RO-MAN) 2018 Abstract [HTML] [PDF]   Conference paper

   Many quadruped robots have been developed to imitate their biological counterparts, several of which show excellent performance. However, the biological neural control mechanisms responsible for self-organized adaptive quadruped locomotion remain elusive. By drawing lessons from biological findings and using an artificial neural approach, we simulated a mammal-like quadruped robot and used it as our simulation platform to investigate and develop neural control mechanisms. In this study, we proposed an adaptive neural control network that can autonomously generate self-organized emergent locomotion with adaptability for the robot. The control network consists of three main components: Decoupled neural central pattern generator circuits (one for each leg), sensory feedback adaptation with dual-rate learning, and multiple neural reflex mechanisms. Simulation results show that the robot can perform quadruped-like gaits in a self-organized manner and adapt its gait to negotiate an obstacle. In addition, this work also suggests that the tight combination of the body-environment interaction and adaptive neural control, guided by sensory feedback adaptation and neural reflexes, is a powerful approach to better understand and solve self-organized adaptive coordination problems in quadruped locomotion.

2. Self-organized quadruped locomotion and body attitude stabilization underadaptive neural control and reflexes Tao Sun, zhendong Dai, Poramate Manoonpong In The 2nd International Youth Conference of Bionic Engineering (IYCBE 2018) 2018 2018 Abstract [HTML]   Conference paper

   The versatile locomotion ability of quadruped animals has been widelyrecognized. Their movements show the impression of elegance includingadaptable, energy-efficient, and self-organized locomotion (Hoyt and Taylor,1981). Many quadruped robots have been developed to imitate their biologicalcounterparts (Raibert, 1986; Izumi et al., 2008; Fukuoka and Kimura, 2009;Wooden et al., 2010; Semini et al., 2011). There have been several biologicalevidences recorded over the years. These biological data identify key ingredientsunderlying locomotion of quadruped animals. The ingredients include centralpattern generators (CPGs), sensory feedback, vestibulospinal reflexes, andstepping reflexes (Boothe, 1999; Kimura et al., 2003; Kimura et al., 2007;Fukuoka and Kimura, 2009). Additionally, biological research found that acombination of CPGs and reflexes are crucial for successful quadruped locomotionin varied terrains (Manoonpong et al., 2007; Fukuoka and Kimura, 2009; Liu etal., 2018). However, quadruped robots’ levels of performance are still far fromthe natural ones. Moreover, the biological neural control mechanisms responsiblefor self-organized quadruped locomotion remain elusive, and the exact way ofhow the combination is achieved remains unknown. In fact, animal self-organized locomotion mechanisms seem to largely dependnot only on CPGs, reflexes, and sensory feedback but also on body-environmentinteraction (Owaki et al., 2012; Barikhan et al., 2014). They interact to generateemerging gaits with adaptability. Generally, CPGs organize basic rhythmicpatterns which are shaped and triggered by sensory feedback through reflexeswhile the body-environment interaction can form adaptive gaits to deal withchanges of the environment. By drawing lessons from the biological concepts andusing a modular artificial neural control approach (Hesse et al., 2012), wesimulated a mammal-like quadruped robot and used it as our experimentalplatform to investigate self-organized quadruped locomotion and body attitudestabilization under adaptive neural control and reflexes. In this study, self-organized locomotion and body attitude stabilization of thesimulated robot are achieved via a sensorimotor loop which involves adaptiveneural control, sensory feedback, and robot body dynamics (Fig. 1(A)). We proposed an adaptive neural control network (Fig. 1(B)) that can autonomouslygenerate self-organized emergent locomotion with adaptability for the robot. Thecontrol network consists of three main components: Decoupled neural centralpattern generator circuits, sensory feedback adaptation with dual-rate learning,and neural reflex mechanisms. Decoupled neural CPG circuits, based on neuraloscillators, produce rhythmic signals which are transmitted to motor neurons.Sensory feedback with adaptive factors and physical body-environmentinteraction adjusts the shapes and phases of the rhythmic signals and, thus,triggers the CPG circuits to form an adaptive gait. Neural reflexes, inspired by thevestibulospinal reflex, act as body attitude control to obtain dynamic stability ofthe robot on a tilting platform. Simulation results show that the robot canperform quadruped-like gaits in a self-organized manner and maintain its bodyattitude stability to deal with different tilting angles of the platform (Fig. 1(C, D)).In addition, this work also suggest that the combination of body-environmentinteraction and adaptive neural control is a powerful approach to solve self-organized quadruped locomotion and body attitude stabilization.