Biological Counterparts
In this tutorial we draw contrasts between biological systems and robot components to help you best understand how the entire neurorobotics system functions. Neurorobotics is a multidisciplinary field that merges neuroscience, robotics, and artificial intelligence. We designed Neurorobotics Studio making it easy to leverage inspirations from how the biological brain controls the body and apply it to the field of robotics. Neurorobotics integrates an artificial brain inspired by a biological brain in the form of neural networks within the robot’s body. Such integration enables a robot to engage and adapt to its environment. The robot's body provides it the ability to interact with its surroundings through sensory perception and motor actions.
Neurorobotics Inspired By Life
In this tutorial we draw contrasts between biological systems and robot components to help you best understand how the entire neurorobotics system functions. Neurorobotics is a multidisciplinary field that merges neuroscience, robotics, and artificial intelligence. We designed Neurorobotics Studio making it easy to leverage inspirations from how the biological brain controls the body and apply it to the field of robotics. Neurorobotics integrates an artificial brain inspired by a biological brain in the form of neural networks within the robot’s body. Such integration enables a robot to engage and adapt to its environment. The robot's body provides it the ability to interact with its surroundings through sensory perception and motor actions.
PNS and CNS
In vertebrates, the Central Nervous System (CNS) and Peripheral Nervous System (PNS) are two primary components of the overall nervous system, each with distinct functions and structures. 1. The PNS sends and receives information between your body and brain. It is composed of sensory receptors, sensory neurons, and motor neurons, which are protected by the vertebral column. These work together to transmit information from and to the central nervous system from designated body parts. In Neurorobotics Studio, it is represented by a robot embodiment’s wheel, camera, etc. 2. The CNS processes and responds to sensory information, acting as the control center for processing sensory information, facilitating cognitive functions, emotions, and decisions. It is composed of the brain and the spinal cord, which are protected by the skull. Together, the brain and spinal cord store, comprehend, and execute based on the information received. The spinal cord acts as a conduit for signals between the brain and the rest of the body, also hosting reflex actions. In Neurorobotics Studio, it is represented by input cortical areas, such as vision, infrared sensors, etc.
FEAGI Implementation of PNS and CNS
In Neurorobotics Studio, a FEAGI brain hosted in a cloud infrastructure represents the Central Nervous System where artificial neurons and synapses enable cognitive functions. Intricate computer networking and computer software simulate the Peripheral Nervous System where sensory information from robots are translated to neuronal activities and routed to the FEAGI brain. Furthermore, artificial brain activities are routed back to the robot and translate back to motor commands. There are many different types of robots, many of which are inspired by life, like humanoids. Neurorobotics takes this concept to another level by mimicking neuronal processing. In Neurorobotics Studio, you will be able to actively see and edit the brain’s sensory, actuation, survival, and learning systems.
Sensory System
Like humans, robots can have 7 senses. These include visual sensing for sight, auditory for hearing, tactile for touch, olfactory for smell, gustatory for taste, vestibular for balance and orientation, and proprioception for position.
Visual Sensory System
In particular, the visual sensory systems help us see the world through our eyes. Eccentricity orients the eyes for vision Blinking rapidly closes and opens the eyes Depth perception allows you to understand the perception of how far away objects are Modulation changes the black part of your eye to adjust light levels. Likewise, it can be translated to robotics to allow the robot see the world as if it were a human. Eccentricity shifts the positioning of the central image for a different focus Blinking temporarily blocks visual stimuli Depth perception encodes the distance measurement signals Modulation expands and contracts the size of the central vision
Auditory Sensory System
The auditory sensory system is responsible for hearing and vocalization. Through our ears, we are able to perceive and interpret sound waves. Through our mouths, we are able to produce sounds. Although this feature is not available in FEAGI at the moment, a robot can have microphones and speakers to receive and emit sounds.
Vestibular and Propioception Sensory System
The vestibular and proprioception sensory systems help with orientation and positioning. To sense the speed that you're traveling at, velocity sensing allows you to detect changes in acceleration through head movements. Body awareness is the body's ability to sense movement, action, and location. Orientation sensing helps us maintain balance and stability by detecting rotational movements. Since robots don't rely on the inner ears' fluids for balance, they rely on features like accelerometers to encode xyz coordinates and gyroscopes to encode pitch, roll, and yaw.
Actuation Sensory System
Unlike the sensory system that interprets sensory information from the peripheral nervous system, the actuation system turns that information into action. Some of the various actuations that our body does on a daily basis include: muscle activation to initiate muscle contraction used to walk forward and backward joint regulation used to execute rotational movements (like a knee bending for a squat) spatial mapping to coordinate movements in multiple directions (like jumping or running diagonally) speed to measure how fast or slow a movement is directional control that allows us to move in yaw, roll, pitch, and cardinal directions. To execute these movements in robots, Neurorobotics Studio uses designated inputs and outputs. Motor outputs moves motors, such as wheels, forward and backward. Servo outputs adjust the degree of rotation, like tilting a head. Navigation instructs specific movements across the x, y, and z axes for moving things like drones diagonally. Speed is used in conjunction with navigation to adjust the fastness or slowness of those movements. Motion control is in charge of yaw, roll, pitch, and cardinal direction movements, which is applicable to embodiments like a plane that can tilt downwards to roll to a side.
Survival Sensory System
Our human body's ultimate goal is to maintain homeostasis for our survival by managing energy, responsiveness, and vitality. To do this, brain activation stimulates brain activity, indicating our brain is responsive. Our metabolism is responsible for processing and utilizing energy through digestion, which is why we feel hungry and eating food gives us energy. If our brain were to shut down, all functions would shut off, leading to a loss of consciousness and vital functions that keep us alive. We are also able to stop ourselves from performing potentially harmful action by suddenly stopping our activity. Even though robots don't have an instinctive urge to remain alive, technology has given them the power to manage their energy, responsiveness, and vitality. In FEAGI: Power acts as an indicator of brain responsiveness and a continuous source of neuronal activation Battery gauge detects the robot's energy stores Death induces a brain death Emergency stop, stops the robot instantly.
Learning Sensory System
To develop our brain, we rely on memory and reinforcement. Through these processes, we learn what desirable behaviors are. Specifically, we rely on 3 types of learning to promote this. Reinforcement learning involves using positive and negative cues, like clapping or booing, to better understand what is right versus wrong. Supervised learning involves receiving cues from external sources, like how a teacher provides you example math problems on a chalkboard. Experiential learning involves experiencing different situations and interactions with one’s environment. Likewise, robots can use reinforcement learning via shocks to imitate punishments. Relating to supervised learning, FEAGI’s ID recognition and ID trainer cortical areas are able associate a neuronal identity to an object. Neuronal plasticity (seen here by the circuit builder’s dashed connector) allows a robot to change and adapt to experiences.
