Experimental validation of constraint mitigation algorithm in underwater robot depth control. Journal Title XX(X):1–13 c The Author(s) 2018 Reprints and permission: sagepub.co.uk/journalsPermissions.nav DOI: 10.1177/ToBeAssigned www.sagepub.com/ Juan Luis Rosendo 1,2 , Benoit Clement 2 and Fabricio Garelli 1 Abstract This work explores both modeling and control of the experimental Ciscrea autonomous underwater vehicle (AUV). A six degree-of-freedom model is presented and validated for turn and emerge/sink maneuvers. Then, a constraint compensating algorithm is proposed based on quasi sliding mode conditioning ideas, and added to a pre-existing inaccessible PD controller in order to improve the overall closed-loop response. By considering actuator constraints, the employed technique allows path following at greater speed than the original controller for a given error tolerance. Experimental results on the so-called Ciscrea underwater robot are presented. Keywords Modeling, AUV, sliding mode control, actuator constraints Introduction The study of the marine environment and commercial activities offshore usually has high cost due to the infrastructure necessary, equipment and skilled personnel. The recent campaigns in the Arctic and Antarctica 1 , the study of seabed 2 , the research in algal blooms and the analysis of the stock in fisheries surveys 3 , applications in the oil and gas sector 4 , among others, prove that the oceans can be successfully explored with robotic probes. The use of autonomous robots for these activities, especially AUV, has powered this kind of research. During these activities, AUVs are exposed to an unknown environment where tasks like infrastructure inspection, patrolling areas, or carrying element are the most common. These duties share a common goal: to follow a pre- established path, as fast as possible and with the minimum possible error. As can be seen, a performance trade-off arises. Indeed, if a sharp path is asked by the guidance function of the robot or if a very quick answer is asked for the path completion, the actuators will reach the maximum allowed allocation. The saturation phenomenon will occur and either error or speed performance metrics will degrade (when saturation occurs, the system behavior is like open loop behavior). Thus, physical constraints should be taken into account as long as demanding control objectives are required. Bibliography presents several cases of study about the saturation problem in autonomous systems 5 . In particular for marine systems, several related works can be found in the literature and we present here some of them. In the works by Campos et al. 6 a PD nonlinear control based on saturation functions with varying parameter for depth and yaw set point regulation and trajectory tracking on an underwater vehicle is proposed. Zheng et al. 7 deal with asymmetric saturation over the actuator of a marine vessel. In their work, a Gaussian error function-based continuous differentiable asymmetric model is employed, for the design of the base control with backstepping technique. Another attempt to AUV application is presented by Steenson et al. 8 where the saturation of the actuator was considered directly in the controller tuning through a MPC design. Moreover, Sarhadi et al. 9 take a simpler approach through a model reference adaptive controller with an anti-windup action, which acts over the input signals of an AUV when saturation arises. These last solutions are valuable and they achieve good results, but in general they require a good model of the 1 CONICET and Universidad Nacional de La Plata, Argentina 2 Lab-STICC, UMR CNRS 6585, ENSTA Bretagne, France Corresponding author: Juan Luis Rosendo, Universidad Nacional de La Plata, C.C.91 (1900), La Plata, Argentina Email: benoit.clement@ensta-bretagne.fr Prepared using sagej.cls [Version: 2016/06/24 v1.10]