To understand how to modulate the behavior of underwater swimmers propelled by multiple appendages, we conducted surge maneuver experiments on our biologically-inspired robot, Madeleine. Robot Madeleine is a self-contained, self-propelled underwater vehicle with onboard processor, sensors and power supply. Madeleine's four flippers, oscillating in pitch, can be independently controlled, allowing us to test the impact of flipper phase on performance. We tested eight gaits, four four-flippered and four two-flippered. Gaits were selected to vary the phase, at either 0 or pi rad, between flippers on one side, producing a fore-aft interaction, or flippers on opposite sides, producing a port-starboard interaction. During rapid starting, top-speed cruising, and powered stopping, the power draw, linear acceleration and position of Madeleine were measured. Four-flippered gaits produced higher peak start accelerations than two, but did so with added power draw. During cruising, peak speeds did not vary by flipper number, but power consumption was double in four flippers compared to that of two flippers. Cost of transport (J N(-1) m(-1)) was lower for two-flippered gaits and compares favorably with that of aquatic tetrapods. Two four-flippered gaits produce the highest surge scope, a measure of the difference in peak forward and reverse acceleration. Thus four flippers produce superior surge behavior but do so at high cost; two flippers serve well for lost-cost cruising.

SUMMARY For early vertebrates, a long-standing hypothesis is that vertebrae evolved as a locomotor adaptation, stiffening the body axis and enhancing swimming performance. While supported by biomechanical data, this hypothesis has not been tested using an evolutionary approach. We did so by extending biomimetic evolutionary analysis (BEA), which builds physical simulations of extinct systems, to include use of autonomous robots as proxies of early vertebrates competing in a forage navigation task. Modeled after free-swimming larvae of sea squirts (Chordata, Urochordata), three robotic tadpoles (`Tadros'), each with a propulsive tail bearing a biomimetic notochord of variable spring stiffness, k (N m-1), searched for, oriented to, and orbited in two dimensions around a light source. Within each of ten generations, we selected for increased swimming speed, U (m s-1) and decreased time to the light source, t (s),average distance from the source, R (m) and wobble maneuvering, W (rad s-2). In software simulation, we coded two quantitative trait loci (QTL) that determine k: bending modulus, E (Nm-2) and length, L (m). Both QTL were mutated during replication, independently assorted during meiosis and, as haploid gametes, entered into the gene pool in proportion to parental fitness. After random mating created three new diploid genotypes, we fabricated three new offspring tails. In the presence of both selection and chance events(mutation, genetic drift), the phenotypic means of this small population evolved. The classic hypothesis was supported in that k was positively correlated (r2=0.40) with navigational prowess, NP, the dimensionless ratio of U to the product of R, t and W. However, the plausible adaptive scenario, even in this simplified system, is more complex, since the remaining variance in NP was correlated with the residuals of R and U taken with respect to k, suggesting that changes in k alone are insufficient to explain the evolution of NP.

Although recent work has demonstrated that modularity can increase evolvability in non-embodied systems, it remains to be seen how the morphologies of embodied agents influences the ability of an evolutionary algorithm to find useful and modular controllers for them. We hypothesize that a modular control system may enable different parts of a robot's body to sense and react to stimuli independently, enabling it to correctly recognize a seemingly novel environment as, in fact, a composition of familiar percepts and thus respond appropriately without need of further evolution. Here we provide evidence that supports this hypothesis: We found that such robots can indeed be evolved if (1) the robot's morphology is evolved along with its controller, (2) the fitness function selects for the desired behavior and (3) also selects for conservative and robust behavior. In addition, we show that if constraints (1) and (3) are relaxed, or structural modularity is selected for directly, the robots have too little or too much modularity and lower evolvability. Thus, we demonstrate a previously unknown relationship between modularity and embodied cognition: evolving morphology and control such that robots exhibit conservative behavior indirectly selects for appropriate modularity and, thus, increased evolvability.

In evolutionary robotics, populations of robots are typically trained in simulation before one or more of them are instantiated as physical robots. However, in order to evolve robust behavior, each robot must be evaluated in multiple environments. If an environment is characterized by f free parameters, each of which can take one of n p features, each robot must be evaluated in all n f p environments to ensure robustness. Here, we show that if the robots are constrained to have modular morphologies and controllers, they only need to be evaluated in n p environments to reach the same level of robustness. This becomes possible because the robots evolve such that each module of the morphology allows the controller to independently recognize a familiar percept in the environment, and each percept corresponds to one of the environmental free parameters. When exposed to a new environment, the robot perceives it as a novel combination of familiar percepts which it can solve without requiring further training. A non-modular morphology and controller however perceives the same environment as a completely novel environment, requiring further training. This acceleration in evolvability -the rate of the evolution of adaptive and robust behavior -suggests that evolutionary robotics may become a scalable approach for automatically creating complex autonomous machines, if the evolution of neural and morphological modularity is taken into account.