Limits...
Distributed recurrent neural forward models with synaptic adaptation and CPG-based control for complex behaviors of walking robots.

Dasgupta S, Goldschmidt D, Wörgötter F, Manoonpong P - Front Neurorobot (2015)

Bottom Line: Biological study has revealed that such complex behaviors are a result of a combination of biomechanics and neural mechanism thus representing the true nature of embodied interactions.Inspired by these findings, we present here, an artificial bio-inspired walking system which effectively combines biomechanics (in terms of the body and leg structures) with the underlying neural mechanisms.Furthermore, we demonstrate that the newly developed recurrent network based approach to online forward models outperforms the adaptive neuron forward models, which have hitherto been the state of the art, to model a subset of similar walking behaviors in walking robots.

View Article: PubMed Central - PubMed

Affiliation: Institute for Physics - Biophysics, George-August-University Göttingen, Germany ; Bernstein Center for Computational Neuroscience, George-August-University Göttingen, Germany ; Laboratory for Neural Computation and Adaptation, Riken Brain Science Institute Saitama, Japan.

ABSTRACT
Walking animals, like stick insects, cockroaches or ants, demonstrate a fascinating range of locomotive abilities and complex behaviors. The locomotive behaviors can consist of a variety of walking patterns along with adaptation that allow the animals to deal with changes in environmental conditions, like uneven terrains, gaps, obstacles etc. Biological study has revealed that such complex behaviors are a result of a combination of biomechanics and neural mechanism thus representing the true nature of embodied interactions. While the biomechanics helps maintain flexibility and sustain a variety of movements, the neural mechanisms generate movements while making appropriate predictions crucial for achieving adaptation. Such predictions or planning ahead can be achieved by way of internal models that are grounded in the overall behavior of the animal. Inspired by these findings, we present here, an artificial bio-inspired walking system which effectively combines biomechanics (in terms of the body and leg structures) with the underlying neural mechanisms. The neural mechanisms consist of (1) central pattern generator based control for generating basic rhythmic patterns and coordinated movements, (2) distributed (at each leg) recurrent neural network based adaptive forward models with efference copies as internal models for sensory predictions and instantaneous state estimations, and (3) searching and elevation control for adapting the movement of an individual leg to deal with different environmental conditions. Using simulations we show that this bio-inspired approach with adaptive internal models allows the walking robot to perform complex locomotive behaviors as observed in insects, including walking on undulated terrains, crossing large gaps, leg damage adaptations, as well as climbing over high obstacles. Furthermore, we demonstrate that the newly developed recurrent network based approach to online forward models outperforms the adaptive neuron forward models, which have hitherto been the state of the art, to model a subset of similar walking behaviors in walking robots.

No MeSH data available.


Related in: MedlinePlus

(A) The closed-loop architectural diagram of an artificial bio-inspired walking system consisting of the sensors (i.e., proprioceptive and exteroceptive sensors) that receive environmental inputs and feedback, the neural mechanisms (i, ii, iii) for adaptive locmotion control, and the biomechanical setup of the hexapod robot AMOSII [i.e., six 3-jointed legs, a segmented body structure with one active backbone joint (BJ), actuators, and passive compliant components Manoonpong et al., 2013]. (B) Modular Robot Control Environment embedded in the LPZRobots simulation toolkit (Der and Martius, 2012; Hesse et al., 2012). (Top left) The simulation environment provides the main testbed for developing the controller, testing it on the simulated hexapod robot, and finally transferring it to the physical agent. Here we evaluate our model and results primarily on the simulated robot (bottom left), which accurately embodies the characteristics of its physical equivalent, AMOS II robot (bottom left). Here, FC1, FC2, FC3, FC4, FC5, and FC6 are foot contact sensors installed in the robot legs, which are used as the main sensory stimuli compared against the predicted signal from the RNN-based (reservoir) forward models. Each leg (bottom right inset) consists of three joints: the innermost thoraco-coxal (TC-) joint enables forward and backward movements, the middle coxa-trochanteral (CTr-) joint enables elevation and depression of the leg, and the outermost femur-tibia (FTi-) joint enables extension and flexion of the tibia. The morphology of these multi-jointed legs were designed based on a cockroach leg (Zill et al., 2004). (Top right) The front and back parts of the body are connected with a backbone joint (BJ) which primarily allows upwards and downwards tilting of the front body segment (along the horizontal axis). Thus, this is used for climbing and gap crossing purposes. This is also based on a similar joint structure found in the cockroach morphology, allowing it to climb large obstacles. More details on BJ control for climbing can be found in Goldschmidt et al. (2014).
© Copyright Policy
Related In: Results  -  Collection

License
getmorefigures.php?uid=PMC4585172&req=5

Figure 1: (A) The closed-loop architectural diagram of an artificial bio-inspired walking system consisting of the sensors (i.e., proprioceptive and exteroceptive sensors) that receive environmental inputs and feedback, the neural mechanisms (i, ii, iii) for adaptive locmotion control, and the biomechanical setup of the hexapod robot AMOSII [i.e., six 3-jointed legs, a segmented body structure with one active backbone joint (BJ), actuators, and passive compliant components Manoonpong et al., 2013]. (B) Modular Robot Control Environment embedded in the LPZRobots simulation toolkit (Der and Martius, 2012; Hesse et al., 2012). (Top left) The simulation environment provides the main testbed for developing the controller, testing it on the simulated hexapod robot, and finally transferring it to the physical agent. Here we evaluate our model and results primarily on the simulated robot (bottom left), which accurately embodies the characteristics of its physical equivalent, AMOS II robot (bottom left). Here, FC1, FC2, FC3, FC4, FC5, and FC6 are foot contact sensors installed in the robot legs, which are used as the main sensory stimuli compared against the predicted signal from the RNN-based (reservoir) forward models. Each leg (bottom right inset) consists of three joints: the innermost thoraco-coxal (TC-) joint enables forward and backward movements, the middle coxa-trochanteral (CTr-) joint enables elevation and depression of the leg, and the outermost femur-tibia (FTi-) joint enables extension and flexion of the tibia. The morphology of these multi-jointed legs were designed based on a cockroach leg (Zill et al., 2004). (Top right) The front and back parts of the body are connected with a backbone joint (BJ) which primarily allows upwards and downwards tilting of the front body segment (along the horizontal axis). Thus, this is used for climbing and gap crossing purposes. This is also based on a similar joint structure found in the cockroach morphology, allowing it to climb large obstacles. More details on BJ control for climbing can be found in Goldschmidt et al. (2014).

Mentions: The neural mechanisms (Figure 1A) for locomotion control, are designed based on a modular architecture, such that, they comprise of, (i) central pattern generator (CPG)-based control, (ii) reservoir-based adaptive forward models, and (iii) searching and elevation action control. The CPG-based control and the searching and elevation control have been previously discussed in detail in Manoonpong et al. (2013), thus here we will only provide a brief overview of these mechanisms, while the reservoir-based adaptive forward models, which forms the main topic of this work, will be presented in detail in the following section.


Distributed recurrent neural forward models with synaptic adaptation and CPG-based control for complex behaviors of walking robots.

Dasgupta S, Goldschmidt D, Wörgötter F, Manoonpong P - Front Neurorobot (2015)

(A) The closed-loop architectural diagram of an artificial bio-inspired walking system consisting of the sensors (i.e., proprioceptive and exteroceptive sensors) that receive environmental inputs and feedback, the neural mechanisms (i, ii, iii) for adaptive locmotion control, and the biomechanical setup of the hexapod robot AMOSII [i.e., six 3-jointed legs, a segmented body structure with one active backbone joint (BJ), actuators, and passive compliant components Manoonpong et al., 2013]. (B) Modular Robot Control Environment embedded in the LPZRobots simulation toolkit (Der and Martius, 2012; Hesse et al., 2012). (Top left) The simulation environment provides the main testbed for developing the controller, testing it on the simulated hexapod robot, and finally transferring it to the physical agent. Here we evaluate our model and results primarily on the simulated robot (bottom left), which accurately embodies the characteristics of its physical equivalent, AMOS II robot (bottom left). Here, FC1, FC2, FC3, FC4, FC5, and FC6 are foot contact sensors installed in the robot legs, which are used as the main sensory stimuli compared against the predicted signal from the RNN-based (reservoir) forward models. Each leg (bottom right inset) consists of three joints: the innermost thoraco-coxal (TC-) joint enables forward and backward movements, the middle coxa-trochanteral (CTr-) joint enables elevation and depression of the leg, and the outermost femur-tibia (FTi-) joint enables extension and flexion of the tibia. The morphology of these multi-jointed legs were designed based on a cockroach leg (Zill et al., 2004). (Top right) The front and back parts of the body are connected with a backbone joint (BJ) which primarily allows upwards and downwards tilting of the front body segment (along the horizontal axis). Thus, this is used for climbing and gap crossing purposes. This is also based on a similar joint structure found in the cockroach morphology, allowing it to climb large obstacles. More details on BJ control for climbing can be found in Goldschmidt et al. (2014).
© Copyright Policy
Related In: Results  -  Collection

License
Show All Figures
getmorefigures.php?uid=PMC4585172&req=5

Figure 1: (A) The closed-loop architectural diagram of an artificial bio-inspired walking system consisting of the sensors (i.e., proprioceptive and exteroceptive sensors) that receive environmental inputs and feedback, the neural mechanisms (i, ii, iii) for adaptive locmotion control, and the biomechanical setup of the hexapod robot AMOSII [i.e., six 3-jointed legs, a segmented body structure with one active backbone joint (BJ), actuators, and passive compliant components Manoonpong et al., 2013]. (B) Modular Robot Control Environment embedded in the LPZRobots simulation toolkit (Der and Martius, 2012; Hesse et al., 2012). (Top left) The simulation environment provides the main testbed for developing the controller, testing it on the simulated hexapod robot, and finally transferring it to the physical agent. Here we evaluate our model and results primarily on the simulated robot (bottom left), which accurately embodies the characteristics of its physical equivalent, AMOS II robot (bottom left). Here, FC1, FC2, FC3, FC4, FC5, and FC6 are foot contact sensors installed in the robot legs, which are used as the main sensory stimuli compared against the predicted signal from the RNN-based (reservoir) forward models. Each leg (bottom right inset) consists of three joints: the innermost thoraco-coxal (TC-) joint enables forward and backward movements, the middle coxa-trochanteral (CTr-) joint enables elevation and depression of the leg, and the outermost femur-tibia (FTi-) joint enables extension and flexion of the tibia. The morphology of these multi-jointed legs were designed based on a cockroach leg (Zill et al., 2004). (Top right) The front and back parts of the body are connected with a backbone joint (BJ) which primarily allows upwards and downwards tilting of the front body segment (along the horizontal axis). Thus, this is used for climbing and gap crossing purposes. This is also based on a similar joint structure found in the cockroach morphology, allowing it to climb large obstacles. More details on BJ control for climbing can be found in Goldschmidt et al. (2014).
Mentions: The neural mechanisms (Figure 1A) for locomotion control, are designed based on a modular architecture, such that, they comprise of, (i) central pattern generator (CPG)-based control, (ii) reservoir-based adaptive forward models, and (iii) searching and elevation action control. The CPG-based control and the searching and elevation control have been previously discussed in detail in Manoonpong et al. (2013), thus here we will only provide a brief overview of these mechanisms, while the reservoir-based adaptive forward models, which forms the main topic of this work, will be presented in detail in the following section.

Bottom Line: Biological study has revealed that such complex behaviors are a result of a combination of biomechanics and neural mechanism thus representing the true nature of embodied interactions.Inspired by these findings, we present here, an artificial bio-inspired walking system which effectively combines biomechanics (in terms of the body and leg structures) with the underlying neural mechanisms.Furthermore, we demonstrate that the newly developed recurrent network based approach to online forward models outperforms the adaptive neuron forward models, which have hitherto been the state of the art, to model a subset of similar walking behaviors in walking robots.

View Article: PubMed Central - PubMed

Affiliation: Institute for Physics - Biophysics, George-August-University Göttingen, Germany ; Bernstein Center for Computational Neuroscience, George-August-University Göttingen, Germany ; Laboratory for Neural Computation and Adaptation, Riken Brain Science Institute Saitama, Japan.

ABSTRACT
Walking animals, like stick insects, cockroaches or ants, demonstrate a fascinating range of locomotive abilities and complex behaviors. The locomotive behaviors can consist of a variety of walking patterns along with adaptation that allow the animals to deal with changes in environmental conditions, like uneven terrains, gaps, obstacles etc. Biological study has revealed that such complex behaviors are a result of a combination of biomechanics and neural mechanism thus representing the true nature of embodied interactions. While the biomechanics helps maintain flexibility and sustain a variety of movements, the neural mechanisms generate movements while making appropriate predictions crucial for achieving adaptation. Such predictions or planning ahead can be achieved by way of internal models that are grounded in the overall behavior of the animal. Inspired by these findings, we present here, an artificial bio-inspired walking system which effectively combines biomechanics (in terms of the body and leg structures) with the underlying neural mechanisms. The neural mechanisms consist of (1) central pattern generator based control for generating basic rhythmic patterns and coordinated movements, (2) distributed (at each leg) recurrent neural network based adaptive forward models with efference copies as internal models for sensory predictions and instantaneous state estimations, and (3) searching and elevation control for adapting the movement of an individual leg to deal with different environmental conditions. Using simulations we show that this bio-inspired approach with adaptive internal models allows the walking robot to perform complex locomotive behaviors as observed in insects, including walking on undulated terrains, crossing large gaps, leg damage adaptations, as well as climbing over high obstacles. Furthermore, we demonstrate that the newly developed recurrent network based approach to online forward models outperforms the adaptive neuron forward models, which have hitherto been the state of the art, to model a subset of similar walking behaviors in walking robots.

No MeSH data available.


Related in: MedlinePlus