Figure 5 - uploaded by Claudio Mattiussi
Content may be subject to copyright.
Components of a microcontroller with von Neumann architecture ~values are given for the Microchip PIC16F268 microcontroller!. The microprocessor unit is composed of an Arithmetic Logic Unit and of control devices to move data from/to memory banks and input/output ports. The memory banks are organized in physically separated locations. For example, the microcontroller shown in the figure uses the ROM memory to store a program composed of a maximum of 2k instructions, a RAM memory to store 224 bytes of data, and an EEPROM memory to store 128 bytes of data. The input/output ports can be connected to sensors, keyboards, LEDs, motorized actuators, or any other peripheral. Gray lines represent the bus where one instruction or data item at a time is moved across components.
Source publication
We describe evolution of spiking neural architectures to control navigation of autonomous mobile robots. Experimental results with simple fitness functions indicate that evolution can rapidly generate spiking circuits capable of navigating in textured environments with simple genetic representations that encode only the presence or absence of synap...
Similar publications
We present a fully-integrated sensing and control system which enables mobile manipulator robots to execute building tasks with millimeter-scale accuracy on building construction sites. The approach leverages multi-modal sensing capabilities for state estimation, tight integration with digital building models, and integrated trajectory planning and...
This paper details the design and implementa-tion of a wheeled mobile robot, which will be referred to as Mobius (Mobile Vision Autonomous System), for self-sustained indoor operation. Its rugged design enables it to be easily customised with auxiliary equipment providing a wide application base. This is facilitated by an accurately controlled high...
Citations
... The weighted connection between sensory input and motor output were optimized by genetic algorithm. Another line of work explores the concept of artificial evolution for development of behavioral abilities without any constrains on the architecture and functioning modalities [28]. Visual information was used as an input to a network of spiking neurons and genetic algorithms were used to evolve it. ...
A considerable amount of research in the field of modern robotics deals with mobile agents and their autonomous operation in unstructured, dynamic, and unpredictable environments. Designing robust controllers that map sensory input to action in order to avoid obstacles remains a challenging task. Several biological concepts are amenable to autonomous navigation and reactive obstacle avoidance. We present an overview of most noteworthy, elaborated, and interesting biologically-inspired approaches for solving the obstacle avoidance problem. We categorize these approaches into three groups: nature inspired optimization, reinforcement learning, and biorobotics. We emphasize the advantages and highlight potential drawbacks of each approach. We also identify the benefits of using biological principles in artificial intelligence in various research areas.
... There has also been extensive work on evolving neural networks in a broad range of domains 29,33,34 ; however, these approaches generally only evolve some small set of (hyper)parameters of the neural network model. For example, they evolve the hyperparamaters of a standard predefined synaptic update rule such as that of STDP [35][36][37][38][39] or predefined spiking models [40][41][42][43][44] . Similarly, reward-based learning has also been evolved but employs the same approach of evolving predefined mathematical models 45,46 . ...
Although deep neural networks have seen great success in recent years through various changes in overall architectures and optimization strategies, their fundamental underlying design remains largely unchanged. Computational neuroscience may provide more biologically realistic models of neural processing mechanisms, but they are still high-level abstractions of empirical behaviour. Here we propose an evolvable neural unit (ENU) that can evolve individual somatic and synaptic compartment models of neurons in a scalable manner. We demonstrate that ENUs can evolve to mimic integrate-and-fire neurons and synaptic spike-timing-dependent plasticity. Furthermore, by constructing a network where an ENU takes the place of each synapse and neuron, we evolve an agent capable of learning to solve a T-maze environment task. This network independently discovers spiking dynamics and reinforcement-type learning rules, opening up a new path towards biologically inspired artificial intelligence.
... In addition to the H-H model, other types of spiking neuron models have been proposed, such as integrate-and-fire models and variants, Izhikevich's neuron model, and spike response model (SRM). Recently, SNN-based models have been applied in variant AI applications, such as character recognition [30,31], object recognition [32], image segmentation [33], speech recognition [34], robotics [35], knowledge representation [36], and symbolic reasoning [37]. In this paper, we will use leaky integrate-and-fire model and Izhikevich's neuron model to convert the word embeddings into more explainable binary embeddings. ...
Word embeddings are the semantic representations of the words. They are derived from large corpus and work well on many natural language tasks, with one downside of costing large memory space. In this paper, we propose binary word embedding models based on inspirations from biological neuron coding mechanisms, converting the spike timing of neurons during specific time intervals into binary codes, reducing the space and speeding up computation. We build three types of models to post-process the original dense word embeddings, namely, the homogeneous Poission processing-based rate coding model, the leaky integrate-and-fire neuron-based model, and the Izhikevich’s neuron-based model. We test our binary embedding models on word similarity and text classification tasks of five public datasets. The experimental results show that the brain-inspired binary word embeddings (which reduce approximately 68.75% of the space) get similar results to original embeddings for word similarity task while better performance than traditional binary embeddings on text classification task.
... The weighted connection between sensory input and motor output were optimized by genetic algorithm. Another line of work explores the concept of artificial evolution for development of behavioral abilities without any constrains on the architecture and functioning modalities [28]. Visual information was used as an input to a network of spiking neurons and genetic algorithms were used to evolve it. ...
... The weighted connection between sensory input and motor output were optimized by genetic algorithm (GA). Another line of work explores the concept of artificial evolution for development of behavioral abilities without any constrains on the architecture and functioning modalities [9]. Vision is used as an input to a network of spiking neurons employing GA to evolve it. ...
... Examples of neuronal networks for control of mobile robots are found in, for example, [Pearson et al., 2007;Floreano et al., 2006;Wang et al., 2008;Johnston et al., 2010]. In particular, in [Pearson et al., 2007] a controller has been designed to be employed on mobile robot vehicles using an FPGA approach, allowing bio-inspired neuronal processing models to be integrated directly into control environments. ...
... To reduce computational cost, relatively simple leaky-integrate-and-fire neurons are used. The neuronal network proposed in [Floreano et al., 2006] uses integrate-and-fire neurons as its constituting units as well. The angular velocities of the motors are determined by the motor neurons, which receive input from excitatory and inhibitory neurons whose response depends on the signals received from sensor neurons. ...
An adaptive training procedure is developed for a network of electronic neurons, which controls a mobile robot driving around in an unknown environment while avoiding obstacles. The neuronal network controls the angular velocity of the wheels of the robot based on the sensor readings. The nodes in the neuronal network controller are clusters of neurons rather than single neurons. The adaptive training procedure ensures that the input–output behavior of the clusters is identical, even though the constituting neurons are nonidentical and have, in isolation, nonidentical responses to the same input. In particular, we let the neurons interact via a diffusive coupling, and the proposed training procedure modifies the diffusion interaction weights such that the neurons behave synchronously with a predefined response. The working principle of the training procedure is experimentally validated and results of an experiment with a mobile robot that is completely autonomously driving in an unknown environment with obstacles are presented.
... A review on bioinspired navigation control schemes is done by Trullier et al. in [16] and by Franz and Mallot [17]. In [18] and [19], Floreano et al. have applied spiking neural circuits to control navigation in a small robot. They used a spiking response model to build a neural network to control the robot. ...
... 8-14. In each of these scenarios, except layouts FIGURE 7. RESISTIVE MEMORY DEVICE MODEL 5 through 7, the robot is required to navigate from the position (1, 1), denoted by '•', to a known target at (19,19), shown as a '×'. The red rectangles are randomly placed obstacles. ...
... 8-14. In each of these scenarios, except layouts FIGURE 7. RESISTIVE MEMORY DEVICE MODEL 5 through 7, the robot is required to navigate from the position (1, 1), denoted by '•', to a known target at (19,19), shown as a '×'. The red rectangles are randomly placed obstacles. ...
Advancement in solid state device technology has made it possible to replicate some learning behaviors on those devices as observed in biological neurons. A very widely used mechanism of learning is Spike Timing Dependent Plasticity (STDP) to realize an unsupervised learning process. In this work, we have developed a novel solution for learning using such networks in robots that can be implemented on novel resistive memory devices. This artificial brain mechanism is capable of learning by observing the environment and taking actions to achieve a desired goal. We have demonstrated this learning scheme using a mathematical model representing the reconfiguration of strengths in synapses. This model can be easily implemented on miniaturized microprocessors using resistive crossbar memories. The reconfigurable resistive memory devices in crossbar arrays are capable of mimicking the synapses in human brains by changing their resistances on application of appropriate voltage signal. In this work, we have demonstrated the potential of this learning scheme by applying it to navigate a two-wheeled differential drive robot in an environment cluttered with obstacles. It can be observed that the robot is able to navigate the environment autonomously, and can reach a given target while avoiding obstacles.
... Trullier et al. [8] and Franz and Mallot [9] have done a review on bioinspired navigation control schemes. Floreano et al. [10] and [11], have applied spiking neural circuits to control navigation in a small robot. They used a spiking response model to build a neural network to control the robot. ...
This paper discusses our work on developing synaptic memory devices based neuromorphic platform for biomimetic robot navigation. We recently developed a Spike Timing Dependent Plasticity (STDP) based learning scheme for robots using mathematical model of memristive devices. We demonstrated the potential of that approach by applying it to navigate a two-wheeled differential drive robot in an environment cluttered with obstacles. In this work, experimentally derived device models for synaptic memory devices are used to extend that learning scheme. Our results indicate that the proposed approach can outperform conventional algorithms in terms of accuracy, realtime unsupervised control, and energy-efficiency.
... SNN-based systems also develop increasingly in the area of robotics, where fast processing is a key issue [104,163,44,43,126,50], from wheels to wings, or legged locomotion. The special abilities of SNNs for fast computing transient temporal patterns make them on first line for designing efficient systems in the area of autonomous robotics. ...
The paper concerns a problem of elaborating a set of the scientific and technical solutions aimed at creating a system of automated quantitative assessment of students’ meta-subject, metacognitive, and meta-creative skills and abilities (meta-competence). It provides overview of free software that implements computer-aided learning methods to be used. There is an emphasis on specialized and universal software. The paper gives recommendations on the choice of tools to implement a pilot sample of a software package. Materials reviewed may also be used to solve a wide range of theoretical and applied tasks regarding the need to implement various computeraided learning methods.
... SNNs can act as brain, controlling and navigating mobile robots. SNNs have at least two properties that make them interesting candidates for adaptive control of autonomous behavioral robots [2,3]. First, the intrinsic dynamic information of SNNs is based on the precision of the spike firing times. ...
... Due to these advantages SNN researchers have tried to employ SNNs in the robotic area. Floreano and colleagues [2,3] proposed a SNN topology based on Genetic Algorithms (GA), and applied it to the Kephera robot controllers for wall following. The GA they used evolved signs (excitation/inhibition) and connections (present/absent) between neurons. ...
Spiking neural network, a computational model which uses spikes to process the information, is good candidate for mobile robot controller. In this paper, we present a novel mechanism for controlling mobile robots based on self-organized spiking neural network (SOSNN) and introduce a method for FPGA implementation of this SOSNN. The spiking neuron we used is Izhikevich model. A key feature of this controller is that it can simulate the process of unconditioned reflex (avoid obstacles using infrared sensor signals) and conditioned reflex (make right choices in multiple T-maze) by spike timing-dependent plasticity (STDP) learning and dopamine-receptor modulation. Experimental results show that the proposed controller is effective and is easy to implement. The FPGA implementation method aims to build up a specific network using generic blocks designed in the MATLAB Simulink environment. The main characteristics of this original solution are: on-chip learning algorithm implementation, high reconfiguration capability, and operation under real time constraints. An extended analysis has been carried out on the hardware resources used to implement the whole SOSNN network, as well as each individual component block.
