Differential mapping spiking neural network for sensor-based robot control

Abstract

In this work, a spiking neural network (SNN) is proposed for approximating differential sensorimotor maps of robotic systems. The computed model is used as a local Jacobian-like projection that relates changes in sensor space to changes in motor space. The SNN consists of an input (sensory) layer and an output (motor) layer connected through plastic synapses, with inter-inhibitory connections at the output layer. Spiking neurons are modeled as Izhikevich neurons with a synaptic learning rule based on spike timing-dependent plasticity. Feedback data from proprioceptive and exteroceptive sensors are encoded and fed into the input layer through a motor babbling process. A guideline for tuning the network parameters is proposed and applied along with the particle swarm optimization technique. Our proposed control architecture takes advantage of biologically plausible tools of an SNN to achieve the target reaching task while minimizing deviations from the desired path, and consequently minimizing the execution time. Thanks to the chosen architecture and optimization of the parameters, the number of neurons and the amount of data required for training are considerably low. The SNN is capable of handling noisy sensor readings to guide the robot movements in real-time. Experimental results are presented to validate the control methodology with a vision-guided robot.

A Fully Spiking Neural Control System Based on Cerebellar Predictive Learning for Sensor-Guided Robots

“…The DM is the differential map [16] needed to generate motor commands in manner similar to the motor cortex, thus acting as an inverse dynamic model, where motor commands ( θcmd ) are generated based on the current states (the spatial velocity ẋs and joint angles θ). In robotics, this is usually expressed in the form of an inverse Jacobian matrix J # DM (θ) which estimates the required motor command to obtain a target spatial velocity at the end effector ẋref .…”

“…The synapses are modulated accordingly to form the differential map, and the SNN in DM can be used to guide the robot in servoing tasks after several iterations [16]. However, the SNN in DM represents a coarse map which lacks in accuracy and precision and needs to be modulated to be adequate for fine motion control.…”

A Bio-Inspired Mechanism for Learning Robot Motion From Mirrored Human Demonstrations

“…• Solving the correspondence issue via SOMs and a biologically inspired plasticity rule. • Improving the performance of a feedforward SNN (Zahra et al, 2021c) relying on Bayesian optimization and inhibitory interconnections. • Validating the improvement in representation capabilities of the developed SNN via complementing the training data.…”

“…The value of the objective function l(γ ) successfully converges to a value of 0.58 rad after around 170 iterations to obtain the values for the network parameters in Table2. This allows to lower the mean value of the maximum deviation error from around 62 mm, following the tuning method introduced inZahra et al (2021c) to 46 mm in the studied workspace and reduction of the number of neurons per neuron assembly from 136 to 20 neurons. It can be noticed in Figure12that spikes occur in the fitness values, which indicates the balance held between exploration and exploitation while searching for the optimal values.SOM: The mapping of the joint-spaces is studied first using the basic SOM developed by Kohonen, as shown in Figure9, to provide a reference value for the improvement in the accuracy of the provided estimations for using the varying density SOM instead, as shown in Figure10.…”

A Computational Framework of Goal Directed Voluntary Motion Generation and Control Loop in Humanoid Robots

“…Thus, developing comprehensive computational frameworks for understanding and approximating the motor behavior system of the human brain is one of the sensible approaches to achieve this. As a natural way, computational modeling of the central nervous system and brain can help for developing a novel motion generation and control system for robots [11][12][13].…”