Designing and controlling a biologically inspired robot fish
Special thanks for my advisers:
Maria-Camilla Fiazza
Madis Listak
Andres Ernits
Chapter 1: Motivation
Traditionally, investigations on fish robots have focused on exploring underwater propulsion. Recently, research has moved towards innovative and biologically – inspired propulsion methods and has also brought direct sensing of environmental flows into the picture. This effort can one day lead to a breakthrough in underwater technology ([1], [2], [3]).
Much fewer investigations explore the idea of interacting with fish using fish robots. The reason is not only the apparent lack of immediate applications but the complexity of the problem; in such an unsteady environment it is hard enough a task to understand and purposefully exploit the relevant laws of physics, realizing realistic locomotion.
The closest achievement so far is a fish robot developed at Essex University [4] remarkably life-like in appearance and capable of replicating a biological swimming pattern during steady locomotion. This robot can swim in a large aquarium tank along with other fish, but it is intended and designed for observation, rather than for interaction.
If we could assume that underwater locomotion had been understood, what questions would we then be able to ask? What would be open for exploration? My thesis is devoted to exploring this question and to technically realizing steps towards its experimental investigation.
In an environment as rich as water, the most diverse and peculiar forms of life exist – it is not a coincidence that the smallest and biggest species are found in the oceans. We humans live on firm ground and have evolved to adapt to our environment; we have explored our environment and learned how to make the most of it, to the point that we have become masters at modifying it to our advantage.
For the longest time, the focus with respect to other species was on how to establish a relationship in terms of resource exploitation. Instead of establishing a symbiosis, we started farming, breeding, raising domestic animals to suit our needs. When we definitively overcame dangerous species and food shortage, it was not a question of survival anymore; so there was room to start observing the animals around us.
Biologists have explored the rich diversity of life forms that nature has developed. Engineers have realized that flawlessly evolved bio-mechanisms hold a tremendous technological potential; an engineering trend towards biologically inspired design has emerged and is now gaining recognition. Biomimetics and biorobotics have been mostly concerned with improvements that can be traced to the study of a single specimen, a representative of its species. However, studying interaction between individuals – either of the same species or across species – can lead us to revolutionary results.
The need to deploy or monitor artificial systems in remote or hostile environments led to the development of underwater vehicles. The propulsive technology used for such vehicles was a port from a mature terrestrial technology, propellers. However, propellers are a technology suitable more for open-water applications than for the cluttered environments in which the richness of fish behaviors can be
fully explored. For example, propellers are a poor choice in coral reefs, near-shore applications, large aquarium tanks and similar niches in which maneuverability is key.
Water as medium for agile motion is still barley explored, for the difficulties that aquatic environments provide. Underwater vehicle design started from rigid metal embodiment with propellers and is now confronting application domains (such as military surveillance applications) in which great maneuverability is needed, along with fast, efficient and silent motion.
Aquatic animals exhibit a wide range of propulsive techniques, from undulatory swimming to lift-based propulsion; furthermore, aquatic animals highlight that there is a key interplay between morphological features and propulsion – ranging from hydrodynamic streamlining to the effective use of fins. There is a lot to learn from the solutions nature has developed.
It should be kept in mind that the features of such natural solutions have co-evolved with the environment, and are thus inextricably linked to the ecological niche; moreover, physiology, morphology and locomotor functionalities have also jointly evolved, in a manner and according to functional principles that are still not completely understood. Consequently, in the design of artificial agents for aquatic environments, the biological reference can only be kept as an inspiration; actual implementation of the robotic device is a matter of adapting the principles, relying on creativity and intuition.
Observing and studying only individual fish would leave many interesting questions unanswered – especially in the case of species that exhibit swarm behavior or in which individuals are highly cross-dependent on other members of their group. For example, schooling behavior has not only evolved as a predator-avoidance strategy, but also as means to enhance locomotor efficiency of theindividual: schooling patterns may be such that individuals benefit from the vortices created by other members of the school [5].
I find it especially interesting to observe fish living in groups and lead by a leader – for example three-spined thicklebacks. It is scientifically challenging to identify the criteria that make a fish the test of its group, as well as to highlight the behavioral or physical features that make a particular specimen most likely to be followed by others. It is an engineering challenge to design a robotic fish (a “carrier” fish) that could successfully lead other fish, by matching their expectation of what a good leader is like.
A carrier robot could help to establish new experimental methods and open up the possibility of communicating to fish what we would like them to do. Scientific experimentation with live fish has always suffered from this difficulty: that we have no way of asking the test specimen to display a particular behavior of interest. We can surely provide food as an incentive to reach a certain location, but we can hardly ask a fish to swim along a predefined trajectory. A carrier robot could deliver the instruction by example – especially if the robot is successfully acknowledged as a leader. Leadership could also be established by rewarding imitation of the carrier. It would then be possible to study which features of the input behavior fish easily replicate, and which features it does not come natural to copy.
Scope
This thesis is devoted to exploring the basic features that such a carrier fish must realize and to implementing a robotic platform that could serve as a carrier. The platform should be designed in such a manner that it could potentially allow investigations into leadership and social roles – especially with respect to being followed. The overall target is designing so that fish can potentially take the robot for one of their own and we can experimentally witness the resulting interaction.
Understanding the features necessary for successful leadership is a very involved task that could take years of research. It is possible that appearance, locomotor abilities, control strategy and sensing abilities all concur to establishing a successful leader – or, on the other hand, it is also possible that only some of these features need to measure up to their biological counterparts. What is required for sure is navigational abilities and good maneuverability.
The complexity of the task implies that intuition will have to take the main role during mechanical design; the design is put to test with a prototype and evaluated qualitatively in terms of smoothness of operation and life-like qualities. Then, ideas for improvement emerge and design can be refined into a successive prototype. We will be designing incrementally for functionality.
I design a biologically inspired robot fish; instead of a rigid body, I implement a novel material combination (silicon-polymer) that leads to a flexible embodiment. However, the operation of a compliant tail is very difficult to model theoretically: not only are compliant bodies non-linear, but also the fish-flow interaction is of an unreasonable complexity to model. Hence, the behavior of the agent can’t be easily deduced from first principles.
A reasonable approach consists in evaluating the forward kinematics to provide an experimental characterization of the prototype. Once it is clear that the prototype possesses suitable locomotor abilities, then it is possible to finalize the control by letting the prototype evolve its own gaits and preferred actuation patterns.
A second approach consists in focusing on sensory-motor coupling and exploring the suitability of the prototype in an experimental environment of controlled complexity, with the robotic equivalent of obstacles, food and predators. Braitenberg vehicles [6] are robots in which the sensor reading control actuation through a per-defined map. The result is rich reactive behavior that is nicely and strongly tied to the environment. The idea captures the fundamental properties of the real world and I believe it provides a smart, effective and elegant solution.
[1] P. Valdivia y Alvarado, Design of Biomimetic Compliant Devices for Locomotion in Liquid Environments, 2007
[2] FILOSE (Robotic FIsh LOcomotion and SEnsing) FP7-ICT-2007-3 STREP research project
[3] IEEE International conference on robotics and automation (ICRA)
[4] http://dces.essex.ac.uk/staff/hhu/
[5] Herskin, J., and SteVensen, J. F.: “Energy savings in sea bass swimming in a school: Measurements of tail beat frequency and oxygen consumption at different swimming speeds.” J. Fish Biol. 53, 366–376., 1998
[6] Braitenberg, V.: “Vehicles: Experiments in synthetic psychology.” Cambridge, MIT, 1984