Adaptive controller for a biomimetic underwater vehicle

Abstract

Aqua is an underwater biomimetic vehicle designed and built at McGill University that uses six oscillating paddles to produce control and propulsion forces. These oscillating paddles provide a time-periodic thrust. Using an existing dynamics model of the vehicle and a numerical simulation, an adaptive controller was developed to provide trajectory tracking capabilities to the vehicle. The performance of the controller was first assessed on a dynamics simulation using different trajectories in roll and pitch. The same controller was then tested experimentally in the Caribbean Sea. We found that the adaptive controller was able to track the roll angle with good accuracy, while the tracking error in the pitch motion was more significant.Résumé : Aqua est un véhicule sous-marin biomimétique conçu et construit à l'Université McGill. Ce véhicule utilise six nageoires oscillantes servant à le contrôler et à le propulser. À l'aide d'un modèle dynamique du véhicule et d'une simulation numérique, une commande adaptative a été élaborée afin de pouvoir suivre une trajectoire désirée du véhicule. La performance de la commande a été évaluée tout d'abord par simulation utilisant différentes trajectoires de roulis et tangage. La même commande a ensuite été testée au cours d'une séance expérimentale dans la mer des Caraïbes. Les résultats ont démontré que la commande adaptative pouvait suivre l'angle de roulis avec bonne précision alors que l'erreur de suivi dans le mouvement de tangage était plus importante. [Traduit par la Rédaction] Mots-clés : contrôles adaptatifs, biomimétique, véhicule sous-marin.

Depth control of the biomimetic U-CAT turtle-like AUV with experiments in real operating conditions

“…For instance, [8] reported good results using PID heading and depth control. In [7], experiments where run in open dynamic environment with adaptive control, reporting satisfactory roll performance, lag in pitch tracking and also concluding that the adaptive law converged to a value similar obtained in the linearization.…”

“…Other examples include Finnegan, the RobotTurtle [4] and iRobot Transiphibian, developed by the same author [5]. Another line of development is represented by AQUA [6] and AQUA2 [7] four finned amphibian robots that are unique in the way the propulsors are used both for swimming and crawling in and out of water. [8] describes a four-finned robot with a controllable fin surface and the authors in particular focus on optimizing the performance of ribbed fins by actively changing the surface contour [9], [10].…”

“…Biomimetic underwater vehicles usually control motion by changing the locomotion primitives of the fins, usually frequency, amplitude, phase shift, or in more complicated cases also the angle of attack, stiffness or surface area. Previously, MPF actuated 4 finned robots have been controlled using adaptive control [7], PID control [8] but also bio-inspired CPG control [14]. U-CAT uses 1 DOF flexible fins but as opposed to the previously cited work, the fins are placed in corners of the vehicle under an angle so that the thrust vectors created by each individual fin are not collinear.…”

Learning legged swimming gaits from experience

“…As mentioned above, accurate modeling of the higherorder dynamics of legged swimming platforms remains an intractable problem given the fluid dynamics involved. However, a number of heuristically useful low-level swimming motions have been determined, through biological inspiration [24] or other modeling techniques. These motions include oscillation, spinning, and braking by statically opposing water flow.…”

“…These effects alone are not enough to simulate the forces generated by an oscillating paddle due to their non-linear dependence on turbulent flow patterns. We use the average thrust model from [24], and provide a set of parameters to tune the effects of paddle motion on the robot's body rotation [3].…”