Octopus behaviour for a novel design

The octopus has fascinated and inspired humans for more then 2000 years. It is a marine invertebrate with amazing motor, sensory and ultimately cognitive capabilities. From an engineering viewpoint its body has interesting characteristics: a soft body devoid of hard skeletal structures, a seemingly infinite number of degrees of freedom (DOFs), bending in many different directions, variable and controllable stiffness, high dexterity, fine manipulation, all managed by a distributed control system. Could the octopus serve as a biological model of how effective behaviour is tightly related to the morphology and physiology of the body? Yes, but one of the main challenges in achieving this goal will be to set a new standard in how closely the robotic and biological scientists collaborate. To reach this goal, learning from each other will be a crucial first step.

In order to facilitate a better understanding I will give a short introduction into how behavioural scientists plan and carry out experiments, using experiments carried out for the OCTOPUS project as an example of what behavioural experiments could offer engineers. We conducted a set of experiments that effected either, bend propagation and fetching movements, or required on-line central control of searching movements and an establishment of learning processes.

In our first experiment we introduced a physical constrain to the base of the octopus arm. Animals were placed inside a transparent Perspex box (40x40x40cm) with a hole at the centre of every surface that allowed the insertion of a single arm only (1.5cm diam.). During the experiment the subjects had to reach out through a hole to retrieve a food reward offered outside the box. The accuracy of the reaching towards a target movement did not improve in consecutive experimental sessions. However, the accuracy and speed of fetching movements improved both within and across sessions. A second set of experiments investigated the ability of octopuses to learn to turn their arm in a specific direction in an opaque Y shaped maze. The animals received neither chemical nor tactile information on the direction of the turn. Therefore the correct decision to turn left or right inside the maze could only be made based on proprioceptive information on the position of the arm. Separate experiments investigated if the octopus can direct its arm inside a maze according to visual feedback. 6 animals learned to direct the movement of their arms inside a three ways choice maze based on visual information.