Lessons for Robotics From the Control Architecture of the Octopus

Abstract

Biological and artificial agents are faced with many of the same computational and mechanical problems, thus strategies evolved in the biological realm can serve as inspiration for robotic development. The octopus in particular represents an attractive model for biologically-inspired robotic design, as has been recognized for the emerging field of soft robotics. Conventional global planning-based approaches to controlling the large number of degrees of freedom in an octopus arm would be computationally intractable. Instead, the octopus appears to exploit a distributed control architecture that enables effective and computationally efficient arm control. Here we will describe the neuroanatomical organization of the octopus peripheral nervous system and discuss how this distributed neural network is specialized for effectively mediating decisions made by the central brain and the continuous actuation of limbs possessing an extremely large number of degrees of freedom. We propose top-down and bottom-up control strategies that we hypothesize the octopus employs in the control of its soft body. We suggest that these strategies can serve as useful elements in the design and development of soft-bodied robotics.

Keywords: biomimetics; neural control architecture; octopus; robotic control; soft robotics.

FIGURE 1

Numbers of neurons and axonal connections in the octopus nervous system. Left: The octopus nervous system is composed of two large populations of neurons, the central brain (optic lobes and brain, 170 million neurons) and peripheral ganglia (350 million neurons) that are connected by orders of magnitude fewer neurons (140 thousand afferent and 32 thousand efferent), creating a bottleneck that requires enormous compression of sensory and motor signals. Right: Numbers for each component of the nervous system on an anatomical diagram.

FIGURE 2

Octopus arm control strategies. Fetching: from the base of the arm, an outbound wave of muscle activation converges with another inbound wave determined by the location of the object (Sumbre et al., 2006). Arm musculature is activated at this midpoint, bending the arm appropriately to pass the object proximally. Sucker recruitment: in response to a stimulus, suckers recruit their neighbors to bend toward this stimulus. These suckers can then recruit their neighbors as this mechanism continues down the arm (Rowell, 1963; Altman, 1968; Gutfreund et al., 2006; Zullo et al., 2011). Arm recruitment: in response to stimulation of one arm, the corresponding suckers on neighboring arms orient toward the site of stimulation (Graziadei, 1965b; Altman, 1968). Grasping: as suckers collectively adhere to an object, sucker recruitment provides multiple afferent pathways for sensory input and multiple efferent pathways for manipulation. If the suckers find prey during foraging, the suckers will recruit their neighbors to capture and immobilize the animal (Rowell, 1963; Altman, 1968; Gutfreund et al., 2006; Zullo et al., 2011). Surface conformation: as suckers recruit their neighbors toward encountered surface features, the arm’s shape conforms to that of the surface (Altman, 1968; Kennedy et al., 2020). Reaching: using visual information the brain determines the horizontal and vertical angle (yaw and pitch) of the arm. The arm then extends by a wave of muscle contraction resembling a propagating bend toward the visual target (Gutfreund et al., 1998; Sumbre et al., 2001; Richter et al., 2015).

FIGURE 3

Hypothetical pipeline of hybrid hierarchical action selection. For each arm, the brain determines an action over a discrete domain (e.g. fetch, reach, push, reject, etc.). For each discrete action, the arm is allowed a subset of continuous stereotyped actions executed based on peripheral proprioceptive information and sensory information from the environment. The actuation of these continuous action subsets overlap within the arm’s configuration space. Most of the arm’s configuration space is dominated by the possible arm shapes resulting from sensory-guided sucker recruitment (e.g. surface conformation). Some behaviors, such as reach, have some continuous parameters that the brain may be able to set (Gutnick et al., 2020).

FIGURE 4

Sucker recruitment used to grasp shrimp meat during foraging. Consistent with localized spread of sucker recruitment, orientation and movement of suckers towards the food occurs in waves propagating from the point of food contact. See also Supplementary Video S1.