Insect Machine Hybrid System


The need of a new approach to understand the adaptability of living organisms

Living organisms exhibit a variety of behaviors such as simple reflexes, innate behaviors, learned behaviors, and cognitive behaviors in complex environments. These behaviors have evolved through interactions between the environment, the nervous system, and the body it controls (Fig. 1). In the adaptive behavior of living organisms, there are some solutions for living organisms to live under circumstances (don't understand meaning of this sentence). Investigating how living organisms receive and process sensory information and how they initiate adaptive behavior in response to such sensory information provide guidelines to understand their adaptability and to construct artificial intelligent systems. So far, the control of behavior in animals has been to a large extent studied with emphasis on brain mechanisms, in particular taking the reductionist approach of analyzing the structure and function of the brain's constituents, individual neurons, and to attempt to reconstruct the system based on such data. However, the brain can only function properly when receiving sensory information about the environment that to some extent also reflects the interaction between the animal and the environment. This fact makes it difficult to understand mechanisms controlling behavior using a bottom-up strategy, despite the fact that it can provide valuable insights. The key problem after building a simulation from the data obtained to emulate mechanisms generating behavior is that assumptions have to be made and the neural mechanisms cannot be confirmed directly. Keeping these shortcomings in mind, a new experimental system is proposed that can complement the bottom-up approach, enable direct tests on adaptive behaviors under controlled conditions, and be utilized in the comparison and verification of models of the nervous system based on the bottom-up approach (Fig 2, 3). In order to meet these various demands, we propose the new approach of an “Insect-machine hybrid system”.

<<<Click here to view the diagram used in the report distributed by AFP (Agence France Press). >>>

Insects as model organisms to understand adaptability

Several millions of animal species, including the human, are living on earth. Of all these animal species, insects are the group that has diversified most impressively, currently representing more than 70% of all animal species and also the majority of terrestrial animal biomass. Odors are distributed in the air in variable patterns, depending factors such as wind direction and speed. Despite such changing environmental conditions, insects are capable of performing odor source localization over a range of several kilometers. Most famously, this has been recounted as an anecdote in Henri Fabre's “Souvenirs entomologiques”. Consider that you depend on a faint odor in the dark in order to localize its source, you will notice how difficult a problem insects are capable to solve. This problem is called chemical plume tracking (CPT) in engineering and is considered to be one of the most difficult technical problems to solve. The odor source orientation behavior of insects is a good example that shows how insects solve problems through the interaction with the environment, modification of behavior, and sensory feedback.

Fig. 1 Brain mechanisms are intimately related to the interaction with the body and the environment.

Male silkmoths (Bombyx mori) initiate an innate behavior, pheromone orientation in response to female sex pheromone. Male silkmoths display a programmed behavioral sequence upon pheromone stimulation consisting of the walking patterns surge, zig-zag and looping. When they preceive a pheromone stimulus again during this behavioral sequence they reset the behavioral pattern and the sequential behavior is released again from its beginning[1]. Odors are distributed in the air in varying distribution patterns. Insects succeed in localizingan odor source by resetting a sequential behavioral pattern repeatedly [1, 10]. The neural circuit releasing this behavior in the silkmoth has been investigated thoroughly from genetic and cellular levels to the whole brain level [1-7]. Besides, the number of neurons constituting the brain amounts only to about 100000 so that the silkmoth can be a excellent model organism to investigate adaptive behaviors of animals at the cellular level. Despite the rigidly programmed behavioral sequence, the threshold and fine-tuning of the behavioral pattern in silkmoth odor source orientation are modulated dynamically by the external environment, such as by visual and odor stimuli and by changes of the internal environment, such as circadian rhythm and experience [2, 12]. Through the interaction with the external environment and changes of the internal environment, the silkmoth is capable of adaptive behavior inchanging environments (Fig.1).

Fig.2 Animal-machine hybrid system

(A)Animal behavior is released through the interaction with body, brain, and environment. (B)Concept for an animal-machine hybrid system.

The purpose of constructing an insect-machine hybrid system

Animals receive a new environmental information when moving and this information is processed in their brains and can be used for motor actions, which in turn affect the relation between the animal and the environment. A brain must display dynamically changing activity patterns resulting from a loop consisting of the environment, the brain itself, and the body of the animal. The adaptive behaviors are generated through these relationships (Fig.1 m2). We propose an “Insect-machine hybrid system” to quantitatively investigate such adaptive behaviors.

In our proposal, consider that the body is replaced by a mobile robot (Fig.2). If the intended movements of the insect can be performed by the robot, the robot can be inserted into the loop consisting of the environment, the brain, and the body. Since the robot can be controlled by the experimenter , the sensory feedback that arises from the intended movements of the insect can be controlled arbitrarily. In other words, the distance and direction of the movement intended by the insect can be faithfully reproduced by the robot or it can be programmed to move according to rules chosen by the experimenter, who can thereby provide the insect with the different, unexpected sensory feedback. If the insect brain can function though the loop formed by the environment, the brain, and the body, we can evaluate adaptive behavior in the insect by determining how the insect can compensate its motor commands in the presence of unexpected sensory feedback. Moreover, it is possible to refine neural circuit models by exposing the insect to systematic perturbations and observing its performance (Fig.3)..
 We implemented two types of insect-machine hybrid systems to meet the demands cited above. The first is an “insect controlled robot” that is controlled by the insect itself by way of its behavior which is recorded; the other is an “insect brain controlled robot” that is controlled by the insect brain.

Fig.3 The purpose of constructing the insect-machine hybrid system
Brain-controlled robot is driven by command signals that transmit from insect brain itself.

The insect brain controlled robot is driven by command signals that the insect brain generates. Consider that the insect brain is replaced by a neural circuit model. If the model can reflect the behavior of the real neural circuit in the insect brain, the output that is generated by the model should correspond to that which is generated by the insect brain. By comparing the model output and the brain output, the evaluation of the model is possible and this is a critical requirement to gradually tune such models to resemble the real brain, with particular emphasis on structure and function.


updating of the site
Copyright (C) 2018 Neuroinformatics Unit, RIKEN Center for Brain Science