A 3D reactive collision avoidance algorithm for underactuated underwater vehicles

Abstract

Avoiding collisions is an essential goal of the control system of autonomous vehicles. This paper presents a reactive algorithm for avoiding obstacles in a threedimensional space, and shows how the algorithm can be applied to an underactuated underwater vehicle. The algorithm is based on maintaining a constant avoidance angle to the obstacle, which ensures that a guaranteed minimum separation distance is achieved. The algorithm can thus be implemented without knowledge of the obstacle shape. The avoidance angle is designed to compensate for obstacle movement, and the flexibility of operating in 3D can be utilized to implement traffic rules or operational constraints. We exemplify this by incorporating safety constraints on the vehicle pitch and by making the vehicle seek to move behind the obstacle, while also minimizing the required control effort. The underactuation of the vehicle induces a sway and heave movement while turning. To avoid uncontrolled gliding into the obstacle, we account for this movement using a Flow frame controller, which controls the direction of the vehicleʼs velocity rather than just the pitch and yaw. We derive conditions under which it is ensured that the resulting maneuver is safe, and these results are verified trough simulations and through full-scale experiments on the Hugin HUS autonomous underwater vehicle. The latter demonstrates the performance of the proposed algorithm when applied to a case with unmodeled disturbances and sensor noise, and shows how the modular nature of the collision avoidance algorithm allows it to be applied on top of a commercial control system. 1 | INTRODUCTION Unmanned vehicles are often intended to operate with limited, delayed, or no human supervision. The environment they operate in can be dynamic or unknown, with only incomplete or partially erroneous a priori information available. This is often the scenario for autonomous underwater vehicles (AUVs), charged with exploration and mapping of the worldʼs oceans. The autonomous nature of AUV missions necessitates the ability to react to the environment, for example, to avoid collision with obstacles. The collision avoidance problem becomes particularly challenging if the vehicle is underactuated, as indeed many AUVs are. An underactuated vehicle is not able to control all degrees of freedom independently, and generally has second-order nonholonomic constraints, making it necessary to include the vehicle dynamics in the design and analysis of the control system (Pettersen & Egeland, 1996). In this paper, we present and analyze a collision avoidance algorithm called the constant avoidance angle algorithm, which we have implemented on a class of underactuated underwater vehicles.

Towards Fully Autonomous UAVs: A Survey

“…A number of geometric-based reactive collision avoidance methods focused on noncooperative scenarios (i.e., dynamic environments) for fixed-wing UAVs or vehicles with nonholonomic constraints adopting the idea of collision cones such as [ 115 , 116 , 117 , 118 , 119 , 120 ]. Many of these approaches use linear or nonlinear guidance laws to align the velocity vector (i.e., controlling heading and flight path angles) in a certain direction while keeping a constant relative distance to the obstacle to avoid collisions.…”

“…A different implementation of collision cones was carried out in [ 120 ] for AUVs; however, the same idea can be applied to UAVs as well. No assumptions were made about the obstacle shape; however, obstacles were modeled as spheres for mathematical development, and it was only assumed that the collision cone to the obstacle can be interpreted from sensor measurements.…”

Advanced Algorithms of Collision Free Navigation and Flocking for Autonomous UAVs

“…A number of geometric-based reactive collision avoidance methods focused on non-cooperative scenarios (i.e. dynamic environments) for fixed-wing UAVs or vehicles with nonholonomic constraints adopting the idea of collision cones such as [113][114][115][116][117][118]. Many of these approaches use linear or nonlinear guidance laws to align the velocity vector (i.e.…”

“…A different implementation of collision cones was done in [118] for AUVs; however, the same idea can be applied to UAVs as well. No assumptions were made about the obstacle shape; however, obstacles were modeled as spheres for the mathematical development, and it was only assumed that the collision cone to the obstacle can be interpreted from sensors measurements.…”

Trajectory tracking control of underactuated tendon‐driven truss‐like manipulator based on type‐1 and interval type‐2 fuzzy logic approach

“…Despite all of the above contributions on structural design and kinematic analysis, this unique and innovative robot currently lacks a mature control method. As with some common underactuated robots, such as underactuated space and underwater robots, [8][9][10] flexible-link robots, 11,12 walking and mobile robots, [13][14][15] and so forth, the control methods mainly include energy-based control, 16,17 sliding mode control, [18][19][20] robust control, [21][22][23] backstepping, 24,25 partial feedback linearization, [26][27][28] and so forth. These approaches are relevant and effective in a type of underactuated systems that have the feature that the actuated and unactuated terms can be separated.…”