The Synthetic Nervous System (SNS) is a biologically inspired neural network (NN). Due to its capability of capturing complex mechanisms underlying neural computation, an SNS model is a candidate for building compact and interpretable NN controllers for robots. Previous work on SNSs has focused on applying the model to the control of legged robots and the design of functional subnetworks (FSNs) to realize dynamical systems. However, the FSN approach has previously relied on the analytical solution of the governing equations, which is difficult for designing more complex NN controllers. Incorporating plasticity into SNSs and using learning algorithms to tune the parameters offers a promising solution for systematic design in this situation. In this paper, we theoretically analyze the computational advantages of SNSs compared with other classical artificial neural networks. We then use learning algorithms to develop compact subnetworks for implementing addition, subtraction, division, and multiplication. We also combine the learning-based methodology with a bioinspired architecture to design an interpretable SNS for the pick-and-place control of a simulated gantry system. Finally, we show that the SNS controller is successfully transferred to a real-world robotic platform without further tuning of the parameters, verifying the effectiveness of our approach.

Morphogenetic gradients are essential to allocate cell fates in embryos of varying sizes within and across closely related species. We previously showed that the maternal NF-κB/Dorsal (Dl) gradient has acquired different shapes in Drosophila species, which result in unequally scaled germ layers along the dorso-ventral axis and the repositioning of the neuroectodermal borders. Here we combined experimentation and mathematical modeling to investigate which factors might have contributed to the fast evolutionary changes of this gradient. To this end, we modified a previously developed model that employs differential equations of the main biochemical interactions of the Toll (Tl) signaling pathway, which regulates Dl nuclear transport. The original model simulations fit well the D. melanogaster wild type, but not mutant conditions. To broaden the applicability of this model and probe evolutionary changes in gradient distributions, we adjusted a set of 19 independent parameters to reproduce three quantified experimental conditions (i.e. Dl levels lowered, nuclear size and density increased or decreased). We next searched for the most relevant parameters that reproduce the species-specific Dl gradients. We show that adjusting parameters relative to morphological traits (i.e. embryo diameter, nuclear size and density) alone is not sufficient to reproduce the species Dl gradients. Since components of the Tl pathway simulated by the model are fast-evolving, we next asked which parameters related to Tl would most effectively reproduce these gradients and identified a particular subset. A sensitivity analysis reveals the existence of nonlinear interactions between the two fast-evolving traits tested above, namely the embryonic morphological changes and Tl pathway components. Our modeling further suggests that distinct Dl gradient shapes observed in closely related melanogaster sub-group lineages may be caused by similar sequence modifications in Tl pathway components, which are in agreement with their phylogenetic relationships.

The Biologically Inspired Robotics group at Case Western Reserve University has developed several innovative designs for a new kind of robot that uses peristalsis, the method of locomotion used by earthworms. Unlike previous wormlike robots, our concept uses a continuously deformable outer mesh that interpolates the body position between discrete actuators. Here, we summarize our progress with this soft hyper-redundant robot.

While soft-bodied animals have an extraordinarily diverse set of robust behaviors, soft-bodied robots have not yet achieved this flexiblity. In this paper, we explore controlling a truly continuously deformable structure with a CPG-like network. Our recently completed soft wormlike robot with a continuously deformable outer mesh, along with a continuum analysis of peristalsis, has suggested the neural control investigated here. We use a Wilson-Cowan neuronal model in a continuum arrangement that mirrors the arrangement of muscles in an earthworm. We show that such a system is well suited to incorporate sensory input and can create both rhythmic and nonrhythmic activity. The system can be controlled using straightforward descending signals whose effects are largely decoupled and can modulate the properties from CPG-like behaviors to static waves. This approach will be useful for designing robotic systems that express multiple adaptive behavioral modes.

We have developed several innovative designs for a new kind of robot that uses peristalsis for locomotion, the same method that earthworms use, and report on the first completed prototype (Fig. 1). This form of locomotion is particularly effective in constrained spaces, and although the motion has been understood for some time, it has rarely been effectively or accurately implemented in a robotic platform. We address some reasons for this, including some common misconceptions within the field. We present a technique using a braided mesh exterior to produce fluid waves of motion along the body of a worm-like robot. We also present a new analytical model of this motion and compare predicted robot velocity to a 2-D simulation. Unlike previous mathematical models of peristaltic motion, our model suggests that friction is not a limiting factor in robot speed, but only in acceleration. The concept is highly scalable, and we present methods of construction at two different scales.

A device that could locomote through curving and tortuous spaces would find many applications in medicine and in industry. Invertebrates such as earthworms and leeches can solve this problem using peristaltic locomotion. We describe a device consisting of three braided pneumatic actuators in series that can successfully locomote peristaltically. The device can locomote forwards and backwards in elevated and curving tubes, and with a plastic sheath around it.

The drawback of these solutions is that one must keep track of the times of occurrences of each spike that initiated the synaptic potentials, and recalculate each exponential in the summation at each time step. This creates a large storage and computational overhead. Since both these equations represent the impulse response of a second-order differential equation, another approach is to numerically integrate additional differential equations for each synapse in the network (Wilson and Bower 1989). We have developed an improved method for computing synaptic conductances that separates equations 1 and 2 into two components: one that is a function of the current time of the simulation and one that accumulates the contributions of previous spike events to the synaptic conductance. We demonstrate that this method requires only the storage of two running sums and the time constants for each synapse, and that it is mathematically equivalent to equations 1 and 2. We will then demonstrate that it is also faster for a given level of precision than numerically integrating differential equations for each synapse. We will first describe our algorithm for equation 1, and then for equation 2.

We present fully distributed neural network architecture for controlling the locomotion of a hexapod robot. The design of this network is directly based on work on the neuroethology of insect locomotion. Previously, we demonstrated in simulation that this controller could generate a continuous range of statically stable insect-like gaits as the activity of a single command neuron was varied and that it was robust to a variety of lesions. We now report that the controller can be utilized to direct the locomotion of an actual six-legged robot, and that it exhibits a range of gaits and degree of robustness in the real world that is quite similar to that observed in simulation.

The robustness of a distributed neural-network controller for locomotion based on insect neurobiology has been used to control a hexapod robot. The robustness of the controller is investigated experimentally. Disabling any single sensor, effector, or central component did not prevent the robot from walking. Furthermore, statically stable gaits could be established using either sensor input or central connections. Thus, a complex interplay between central neural elements and sensor inputs is responsible for the robustness of the controller and its ability to generate a continuous range of gaits. These results suggest that biologically inspired neural-network controllers may be a robust method for robotic control.>