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

Real-time data for adaptive locomotion to overcome leg damage. (A) The FT-i joint angles of the right middle leg R2. (B) The CT-i joint angles of the right middle leg R2. (C) The TC-i joint angles of the right middle leg. (D) Accumulated error signal at the end of each stand phase. It is reset to zero at every swing phase. Below pictures show the locomotion of AMOSII during the experiment (temporal spacing of the panels are not exact). Please see the Supplementary Video 6 for closer look at the exact adaptive behavior.
© Copyright Policy
Related In: Results  -  Collection

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

Figure 9: Real-time data for adaptive locomotion to overcome leg damage. (A) The FT-i joint angles of the right middle leg R2. (B) The CT-i joint angles of the right middle leg R2. (C) The TC-i joint angles of the right middle leg. (D) Accumulated error signal at the end of each stand phase. It is reset to zero at every swing phase. Below pictures show the locomotion of AMOSII during the experiment (temporal spacing of the panels are not exact). Please see the Supplementary Video 6 for closer look at the exact adaptive behavior.

Mentions: As observed in Figure 9, initially AMOSII walks under normal conditions (photo panel 1) with the right middle leg FT-i joint functioning normally. The FT-i joint was then constrained to 0° maximum and minimum angle of clearance (Figure 9A) thereby causing the right middle leg to be suspended in the air (photo panel 2). As a result the reservoir forward model prediction mismatches the current footcontact signal on the damaged leg, causing the accumulated error to gradually ramp up (Figure 9D). After a short transient period of AMOSII trying to walk in this configuration (dark green section in Figure 9), this results in adaptations in the FT-i and CT-i joints (yellow highlighted section in Figures 9A,B) thereby, allowing the robot to extend the damaged leg further down and support the locomotion (photo panels 3, 4, and 5). As a result, AMOSII was able to successfully keep walking straight with a slightly modified tetrapod gait despite the damaged right middle leg. Finally, after 2000 time steps (≈74 s), the FT-i joint was once again allowed to function normally, causing the accumulated error to become zero (the forward model prediction matches the actual footcontact signal). The robot then continues to walk as in the undamaged condition with a tetrapod gait. For further details, we encourage the readers to see the Supplementary Video 6 of the entire experiment. These results, thus clearly demonstrate that the distributed reservoir forward models not only allow complex locomotive behaviors, but also enable the robot to deal with unwanted changes in body properties in a robust manner.


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)

Real-time data for adaptive locomotion to overcome leg damage. (A) The FT-i joint angles of the right middle leg R2. (B) The CT-i joint angles of the right middle leg R2. (C) The TC-i joint angles of the right middle leg. (D) Accumulated error signal at the end of each stand phase. It is reset to zero at every swing phase. Below pictures show the locomotion of AMOSII during the experiment (temporal spacing of the panels are not exact). Please see the Supplementary Video 6 for closer look at the exact adaptive behavior.
© Copyright Policy
Related In: Results  -  Collection

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

Figure 9: Real-time data for adaptive locomotion to overcome leg damage. (A) The FT-i joint angles of the right middle leg R2. (B) The CT-i joint angles of the right middle leg R2. (C) The TC-i joint angles of the right middle leg. (D) Accumulated error signal at the end of each stand phase. It is reset to zero at every swing phase. Below pictures show the locomotion of AMOSII during the experiment (temporal spacing of the panels are not exact). Please see the Supplementary Video 6 for closer look at the exact adaptive behavior.
Mentions: As observed in Figure 9, initially AMOSII walks under normal conditions (photo panel 1) with the right middle leg FT-i joint functioning normally. The FT-i joint was then constrained to 0° maximum and minimum angle of clearance (Figure 9A) thereby causing the right middle leg to be suspended in the air (photo panel 2). As a result the reservoir forward model prediction mismatches the current footcontact signal on the damaged leg, causing the accumulated error to gradually ramp up (Figure 9D). After a short transient period of AMOSII trying to walk in this configuration (dark green section in Figure 9), this results in adaptations in the FT-i and CT-i joints (yellow highlighted section in Figures 9A,B) thereby, allowing the robot to extend the damaged leg further down and support the locomotion (photo panels 3, 4, and 5). As a result, AMOSII was able to successfully keep walking straight with a slightly modified tetrapod gait despite the damaged right middle leg. Finally, after 2000 time steps (≈74 s), the FT-i joint was once again allowed to function normally, causing the accumulated error to become zero (the forward model prediction matches the actual footcontact signal). The robot then continues to walk as in the undamaged condition with a tetrapod gait. For further details, we encourage the readers to see the Supplementary Video 6 of the entire experiment. These results, thus clearly demonstrate that the distributed reservoir forward models not only allow complex locomotive behaviors, but also enable the robot to deal with unwanted changes in body properties in a robust manner.

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