150
C. Goldsmith et al.
behavior. Therefore, it does not appear that combinations involving inhibiting
the extension neurons result in reflex reversal.
The most drastic change to joint motion occurs when inhibiting the flexion
position neurons. The joint’s resulting extension is less than half of that of the
normal case, while the post-stimulus return to equilibrium is much more rapid.
This speed results in a large degree of overshoot into flexion, which is in some
cases briefly sustained. When combined with some degree of velocity inhibition,
the effect becomes more pronounced, with the cases involving high inhibition of
flexion position and velocity resulting in no reaction to joint flexion.
This open loop response resembles motion characterizing the active reaction
(AR), allowing the joint to flex and even assisting in the motion. Figure 6a
provides a closer look at the activity of the network during these responses.
Even in the closed loop case the joint exhibits similar AR behavior, as the sensory
inputs to the NSI - and eventually the MNs - are greatly inhibited throughout
flexion. This data supports the conclusion that inhibiting the flexion position and
velocity sensory neurons is what causes reflex reversal [7]. It is also interesting to
note that the system still produces a resistance reflex against unintended joint
extension (Fig. 6b). This seems to suggest that sensory afferents does not simply
reverse a reflex; instead, it fundamentally changes the operation of the joint,
allowing and assisting motions in the intended direction, while preventing those
in the unintended direction.
4 Discussion
In this manuscript, we presented a simulated neuromechanical model of an insect
FTi joint based on previously observed extension networks in the animal. We
performed experiments on the network by selectively inhibiting groups of sensory neurons and observing the joint’s response. Our findings suggest that by
inhibiting the flexion position sensory neurons, we can cause a transition from
a resistance reflex (RR) to an active reaction (AR) response to fCO elongation.
Simultaneously inhibiting the flexion velocity neurons then works to amplify this
effect, in many cases preventing the flexion neurons from firing during flexion.
Our findings in the inhibited case support the hypothesis that the nervous
system may actively modulate the lowest level sensory afferents in order to alter
motion and control [7,11]. Such an approach seems counter intuitive from a
robotics perspective, since inhibiting sensors means depriving the entire control
system of that information. An engineer might attempt to implement this reflex
reversal by collecting the same sensory information no matter the context and
formulating context-dependent motor commands in response. This way, no sensory information is “lost.” Why, then, would the animal nervous system utilize
the strategy it does? The distributed nature of the nervous system may necessitate it, as the primary role of the sensory information encoded by the afferents
is to execute reflexes for the single joint. Such distributed processing may then
be necessary for fast response times and producing capable locomotion.
Additionally, our findings from the inhibited joint’s response to extension
poses questions about the manner in which the nervous system dictates joint
C. Goldsmith et al.
behavior. Therefore, it does not appear that combinations involving inhibiting
the extension neurons result in reflex reversal.
The most drastic change to joint motion occurs when inhibiting the flexion
position neurons. The joint’s resulting extension is less than half of that of the
normal case, while the post-stimulus return to equilibrium is much more rapid.
This speed results in a large degree of overshoot into flexion, which is in some
cases briefly sustained. When combined with some degree of velocity inhibition,
the effect becomes more pronounced, with the cases involving high inhibition of
flexion position and velocity resulting in no reaction to joint flexion.
This open loop response resembles motion characterizing the active reaction
(AR), allowing the joint to flex and even assisting in the motion. Figure 6a
provides a closer look at the activity of the network during these responses.
Even in the closed loop case the joint exhibits similar AR behavior, as the sensory
inputs to the NSI - and eventually the MNs - are greatly inhibited throughout
flexion. This data supports the conclusion that inhibiting the flexion position and
velocity sensory neurons is what causes reflex reversal [7]. It is also interesting to
note that the system still produces a resistance reflex against unintended joint
extension (Fig. 6b). This seems to suggest that sensory afferents does not simply
reverse a reflex; instead, it fundamentally changes the operation of the joint,
allowing and assisting motions in the intended direction, while preventing those
in the unintended direction.
4 Discussion
In this manuscript, we presented a simulated neuromechanical model of an insect
FTi joint based on previously observed extension networks in the animal. We
performed experiments on the network by selectively inhibiting groups of sensory neurons and observing the joint’s response. Our findings suggest that by
inhibiting the flexion position sensory neurons, we can cause a transition from
a resistance reflex (RR) to an active reaction (AR) response to fCO elongation.
Simultaneously inhibiting the flexion velocity neurons then works to amplify this
effect, in many cases preventing the flexion neurons from firing during flexion.
Our findings in the inhibited case support the hypothesis that the nervous
system may actively modulate the lowest level sensory afferents in order to alter
motion and control [7,11]. Such an approach seems counter intuitive from a
robotics perspective, since inhibiting sensors means depriving the entire control
system of that information. An engineer might attempt to implement this reflex
reversal by collecting the same sensory information no matter the context and
formulating context-dependent motor commands in response. This way, no sensory information is “lost.” Why, then, would the animal nervous system utilize
the strategy it does? The distributed nature of the nervous system may necessitate it, as the primary role of the sensory information encoded by the afferents
is to execute reflexes for the single joint. Such distributed processing may then
be necessary for fast response times and producing capable locomotion.
Additionally, our findings from the inhibited joint’s response to extension
poses questions about the manner in which the nervous system dictates joint