LEGO 로봇에 신경 시스템 시뮬레이션을 설계 및 구현

The overall goal of this method is to quickly prototype and test nervous system simulations on a robotic platform. This is achieved by running neural network simulations on the Lego Mindstorms NXT platform, and by performing side-by-side comparison testing with the animal model. First, a robot is constructed to be used as a platform for the nervous system simulation.

A hypothetical neural network based on neuro ethological study is then developed and programmed onto the robotic platform. Ethological studies are conducted to compare the hypothetical neural networks performance to that of the animal nervous system. Deficiencies in the model can be identified and inform its further development systems level neuroscience hypotheses can be tested by direct manipulation of simulation parameters.

Hello, I'm Dan Bluestein from the Marine Science Center at Northeastern University. Here we'll demonstrate an approach to rapidly implement nervous system simulations on a robotic platform. By developing this procedure for use on a commercially available robotics kit, we have made this bio robotic approach to neuroscience affordable, efficient, and accessible to researchers without robotic engineering experience.

This approach is also well-suited for the high school and college classroom to serve as the foundation for a hands-on inquiry-based bio robotics curriculum. The first step in building the robot platform is to choose a model organism to study that is well represented in the neuro ethological literature. Invertebrates generally make good candidates because their relatively simple nervous systems have been well-studied and are primarily comprised of innate reflexes.

We will demonstrate this approach using the American Lobster Hamas Americana. The next step is to select well-studied reflexive behaviors for modeling purposes. With the lobster, we can investigate responses to inten, bend, claw bump, and optical flow.

Simple reflexes relying on decking neural connections from bilaterally. Symmetrical sensors are well-suited for this type of study. Using the Lego Mindstorms NXT 2.0 kit build a robot that approximates the body structure of the animal model.

Other robot platforms can be used, but the commercially available kit we use here allows for rapid implementation. The next step is to equip the robot with sensors that match the systems chosen for study sensors. Included in the Lego Mindstorms kit can be used or home brew sensors can be engineered as outlined in a variety of published material.

We use the Lego touch sensor, a custom antenna Flex point bend sensor, and an optical flow sensor developed by Centi Incorporated to build a custom resistive connector such as the antenna sensor splice an NXT connector wire and solder the black and white wires to the leads of the sensor. To begin Programming the nervous system simulation use previously published neuro ethological studies to develop a neural network for modeling purposes. The literature has proposed neural networks to explain the behavior of various model organisms from lobsters to lampre, functional neural units should be identified and their synaptic connections.

Theorized novel neuro ethological experiments can also be developed and implemented if laboratory circumstances allow. Use LabVIEW software to program the network for the Lego Mindstorms platform. Virtual instruments or BIS for a neuron and synapse model are used and can be found online.

We recommend the discreet time map-based model developed by Roloff because it allows for computationally efficient real-time operation while maintaining a variety of neuronal firing output regimes two control parameters. Alpha and Sigma determine the dynamics of the simulated neuron to produce varied outputs, including tonic, spiking, chaotic firing, bursting, and intrinsic silence. Lab view provides a graphical programming environment that is intuitive and allows for rapid deployment first place neuron and synapse vi's into a while loop so that the network will update iteratively.

Next, connect pre-synaptic neuron spike outputs through synapses to post synaptic neurons. Then pass iteratively updating parameters through shift registers to the next calculation cycle. The DTM model updates fast and slow dynamical variables that are passed to the next iteration of calculations.

Synapse parameters must be set to define network geometry. The DTM model specifies the strength of a synapse, the relaxation rate and the reversal potential values can be tuned for your specific network characteristics, but we recommend the following settings. JS equals 0.60 gamma equals 0.95 and XRP equals 2.2 for an excitatory synapse or negative 1.8.

For an inhibitory synapse, adjust the alpha and sigma control parameters to modify baseline neuron activity. We recommend using an alpha value equal to 4.05 and a sigma value equal to negative 3.10. Use lab view sensor vs.

To input sensor information into your network. For provided sensors, the vi's are already available within the Mindstorms lab view module. For custom resistant sensors, use the light sensor VI as a programming framework.

Insert graphical charts on the front panel of the lab view program to visualize neuronal activity. Temporarily replace sensor inputs with front panel control boxes in order to manually manipulate the sensor information going to the neural network, run the neuronal network and manually adjust the front panel controls to present varied sensory input. Confirm that the network qualitatively functions as predicted for known conditions.

Adjust the network as needed by changing the neuron and synapse parameters. First attempts at tuning the network should be done by changing the strength of the various synapses. Other parameters may need to be adjusted as well.

Once the network functions with qualitative accuracy, replace the front panel control boxes with the sensor input VI'S program the network onto the NXT platform through a USB connection. To test the nervous system simulation, set up a controlled environment in which to observe the animal and the robot under similar conditions. Mount a video camera overhead to record behaviors.

Make sure that the contrast between the subject and the background is high. To allow for automatic tracking. Place the animal in the controlled environment and video record the behavior.

Be sure to limit the sensory capabilities of the lobster to match those of the robot. Here we restrict the lobster's vision and chemo detection by covering the eyes with an aluminum foil mask and denting chemo receptors with fresh water. Modify the environment to allow for operation of the Lego robot.

Repeat the behavior testing and videotaping procedure using the robot. Use MATLAB to automatically track the animal and the robot. We use the mouse lab tracker MATLAB script, which is available online.

Create and compare plots of the animal and robot movements. Modify the neural network and repeat the behavioral tracking experiment to observe how different aspects of the electronic nervous system affect the robot's behavior. For example, synaptic strengths at certain nodes of the neural network can be adjusted.

Alternatively, the sensory conditions in the test arena can be modified and the effects observed in robot and animal. By comparing the behavioral performance of a robot and an animal, we can investigate principles of nervous system function. Synaptic strength from the claw bum sensory neurons into the nervous system was varied on a robot and the resulting behavior was Tracked.

Red lines represent the movement robots with weak excitatory synapses. Blue lines show medium strength excitatory synapses, and green lines show robot movement with strong excitatory synapses. Lobster paths are represented by black lines, the hypothetical nervous system with mid-level excitatory connections from the claw bump.

Sensory neurons produce similar behavioral output to the animal. Quantitative parameters for comparison can be extracted from the data such as path length and average walking speed. As shown in this table, We have just shown How to use an affordable and accessible robot platform to run and test nervous system simulations.

This approach is useful for the rapid testing of nervous system hypotheses and can be quickly adopted by biologists and other researchers without robotics experience. There are a few inherent limitations in using this platform, such as the processing capabilities of the NXT chip and the range of sensors available for robotic implementation. Nonetheless, this approach is a powerful tool for early stage implementation of embodied nervous system simulations and can serve as a starting point for more in-depth bio robotic investigations using custom platforms when used by high school and college students.

This approach provides an exciting inquiry-based method to teach neuroscience, robotics, and the value of interdisciplinary science to the scientists of the future. Thanks For watching.