SPARK programs robots with insect perception

We talked recently about the importance of complementing robotic perception with some basic cognitive tools, in order to produce robots that can interact with their environment in a more natural fashion. The EU-funded SPARK project has developed a perceptual control architecture that allows robots to adapt to cluttered and changing environments, and learn from experience – an ability that may eventually suit it to disaster or war zones. The design is based on the neural processes of insects, and allows robots to form a unique, abstract representation of the environment, and modulate their behavior accordingly.

Using cutting edge technology to create robo-bugs and replicate the adaptive behavior of insects would be an huge milestone in Artificial Intelligence research. After all, flies may not share our intelligence, but they’ve proved time and again that their evasive airborne maneuvering could put the Red Baron to shame. The ability of insects to maneuver so effectively in their environment stems from their cognitive structure. As insects receive stimuli from their senses, they are presented with a variety of reflexive options. The higher brain functions of the insect organize the information into a picture of the environment, and self-modulate the basic actions that were triggered by the sensory-motor pathways.

This two-fold system of environmental interaction, in which the senses dictate basic response and the higher cognitive systems direct the behavior, was the basis of SPARK’s spatial-temporal array computer based structure (SPARC). The perception architecture allows robots to form, learn and remember navigation strategies without human intervention. SPARK-based robots are fitted with sophisticated tactile, aural and visual sensors, and programmed with basic reactions towards certain stimuli - for example, robots are directed to move themselves towards a specific sound source. However, the reflexive action is regulated by the robot, so that it can move efficiently through a changing environment, towards the sound source, without trapping itself. The ability of the robot to modulate its behavior allows it to learn from experience, and improvise.

“The SPARC architecture is a starting step toward emulating the essential perception-action architecture of living beings, where some basic behaviors are inherited, like escaping or feeding, while others are incrementally learned, leading to the emergence of higher cognitive abilities,” notes Paolo Arena, the project coordinator. “The robot will initially behave by using primarily the basic inherited behaviors,” says Arena. “Higher knowledge will be incrementally formed in the higher layer of the architecture, which is a neuron lattice based on the Reaction-Diffusion Cellular Non-linear Network (RD-CNN) paradigm, able to generate self-organizing dynamic patterns.”

The potential applications for robots that can efficiently navigate through changing, dangerous environments are huge. If robots were programmed to reflexively move toward the aural stimulus of the human voice, and given the cognitive ability to seek a safe pathway through a changing environment, they could be of great use in search and rescue operations in dangerous or inaccessible environments.

In order to develop the SPARC system, the SPARK research team investigated how visual, aural, and tactile senses create a spatial-temporal simplified dynamic representation of the environment. They then used this understanding of perception to create a pattern formation system, in which the “pattern” is the collected sensory data. This framework was then applied to robotic models, which underwent environmental tests, the video results of which can be seen here. Research into insect brain neurobiology is continuing with the SPARK II project, which aims to refine, assess, and generalize the SPARK architecture.