352
N. S. Szczecinski et al.
loading has discrete effects upon CS encoding, specifically, that pre-loading the leg resets
the amplitude sensitivity, while dynamic properties (e.g. encoding the rate of change of
force) are not altered by history [2]. Conversely, the model could be inverted to infer the
instantaneous force acting on a leg given the CS recording during motion. Using animal
kinematic and force measurements to build a model of insect walking has already led to
a better understanding of the types of forces these sensors are subjected to as the animal
walks freely [16]. Better understanding the responses to these forces will elucidate what
information the nervous system has available to it regarding forces applied to its legs.
This model will also benefit robotics. To better understand how the insect nervous
system uses CS feedback in the control of walking, we have built such strain sensors into
the legs of our robots [10, 11]. Such sensors are particularly useful for detecting when a
leg is in the “stance phase,” during which it supports and propels the body, versus when it
is in the “swing phase.” In the past, these sensors have provided our stepping controllers
with non-adapting feedback proportional to the force on the leg. However, calibrating such sensors to eliminate constant offsets while maintaining maximal sensitivity is
critical for proper function; if the offset is too high, the sensors return false-positive
information about leg loading; if the offset is too low, the sensors return false-negative
information. We believe that our CS response model could be implemented to run in realtime onboard robots, enabling their sensors to self-calibrate. Such an algorithm would
adapt to cancel out offsets, but remain sensitive to sudden changes in the force level (e.g.
from a leg transitioning from the swing phase to the stance phase). Such self-calibration
may increase the reliability of large arrays of analog sensors onboard robots that provide
feedback regarding support and contact forces, environmental fluid currents (e.g. via
hairs), and other body-wide conditions.
What specific structures might give rise to the dynamics we describe in this
manuscript? Experimental data and computational modeling of spider mechanoreceptors
suggest that adaptation arises due to adaptive ion channels present in receptor cells [19].
The viscoelastic hysteresis of the exoskeleton and the CS themselves is also known to
contribute to sensory adaptation [20]. Future experiments may reveal additional sources
of adaptation. Better understanding such sources may suggest new sensor designs or
processing algorithms that would endow walking robots with animal-like mobility.
References
1. Zill, S.N., Schmitz, J., Büschges, A.: Load sensing and control of posture and locomotion.
Arthropod Struct. Dev. 33, 273–286 (2004)
2. Ridgel, A.L., Frazier, S.F., DiCaprio, R.A., Zill, S.N.: Encoding of forces by cockroach tibial
campaniform sensilla: implications in dynamic control of posture and locomotion. J. Comp.
Physiol. A Sens. Neural, Behav. Physiol. 186, 359 (2000)
3. Zill, S.N., Büschges, A., Schmitz, J.: Encoding of force increases and decreases by tibial
campaniform sensilla in the stick insect, Carausius morosus. J. Comp. Physiol. A Neuroethol.
Sens. Neural, Behav. Physiol. 197, 851–867 (2011)
4. Zill, S.N., Dallmann, C.J., Büschges, A., Chaudhry, S., Schmitz, J.: Force dynamics and
synergist muscle activation in stick insects: the effects of using joint torques as mechanical
stimuli. J. Neurophysiol. 120, 1807–1823 (2018)
N. S. Szczecinski et al.
loading has discrete effects upon CS encoding, specifically, that pre-loading the leg resets
the amplitude sensitivity, while dynamic properties (e.g. encoding the rate of change of
force) are not altered by history [2]. Conversely, the model could be inverted to infer the
instantaneous force acting on a leg given the CS recording during motion. Using animal
kinematic and force measurements to build a model of insect walking has already led to
a better understanding of the types of forces these sensors are subjected to as the animal
walks freely [16]. Better understanding the responses to these forces will elucidate what
information the nervous system has available to it regarding forces applied to its legs.
This model will also benefit robotics. To better understand how the insect nervous
system uses CS feedback in the control of walking, we have built such strain sensors into
the legs of our robots [10, 11]. Such sensors are particularly useful for detecting when a
leg is in the “stance phase,” during which it supports and propels the body, versus when it
is in the “swing phase.” In the past, these sensors have provided our stepping controllers
with non-adapting feedback proportional to the force on the leg. However, calibrating such sensors to eliminate constant offsets while maintaining maximal sensitivity is
critical for proper function; if the offset is too high, the sensors return false-positive
information about leg loading; if the offset is too low, the sensors return false-negative
information. We believe that our CS response model could be implemented to run in realtime onboard robots, enabling their sensors to self-calibrate. Such an algorithm would
adapt to cancel out offsets, but remain sensitive to sudden changes in the force level (e.g.
from a leg transitioning from the swing phase to the stance phase). Such self-calibration
may increase the reliability of large arrays of analog sensors onboard robots that provide
feedback regarding support and contact forces, environmental fluid currents (e.g. via
hairs), and other body-wide conditions.
What specific structures might give rise to the dynamics we describe in this
manuscript? Experimental data and computational modeling of spider mechanoreceptors
suggest that adaptation arises due to adaptive ion channels present in receptor cells [19].
The viscoelastic hysteresis of the exoskeleton and the CS themselves is also known to
contribute to sensory adaptation [20]. Future experiments may reveal additional sources
of adaptation. Better understanding such sources may suggest new sensor designs or
processing algorithms that would endow walking robots with animal-like mobility.
References
1. Zill, S.N., Schmitz, J., Büschges, A.: Load sensing and control of posture and locomotion.
Arthropod Struct. Dev. 33, 273–286 (2004)
2. Ridgel, A.L., Frazier, S.F., DiCaprio, R.A., Zill, S.N.: Encoding of forces by cockroach tibial
campaniform sensilla: implications in dynamic control of posture and locomotion. J. Comp.
Physiol. A Sens. Neural, Behav. Physiol. 186, 359 (2000)
3. Zill, S.N., Büschges, A., Schmitz, J.: Encoding of force increases and decreases by tibial
campaniform sensilla in the stick insect, Carausius morosus. J. Comp. Physiol. A Neuroethol.
Sens. Neural, Behav. Physiol. 197, 851–867 (2011)
4. Zill, S.N., Dallmann, C.J., Büschges, A., Chaudhry, S., Schmitz, J.: Force dynamics and
synergist muscle activation in stick insects: the effects of using joint torques as mechanical
stimuli. J. Neurophysiol. 120, 1807–1823 (2018)