Session: 07-11-01 Mobile Robots and Unmanned Ground Vehicles I

Paper Number: 70489

Start Time: Tuesday, 02:00 PM

70489 - A Special Strategy for Quadrotors in Aggressive and Singularity-Free Tracking 

Unmanned aerial vehicles (UAVs) face many challenges when navigating through different terrains and obstacles. Quadrotors are a popular type of UAVs, which have been extensively researched in the recent years. Quadrotors, while exhibiting high agility, generally lack stability. To improve the performance and to expand the capability of a quadrotor, it is necessary for the quadrotor to undertake aggressive maneuvers. During an aggressive maneuver process, a traditional quadrotor model based on Euler angles may encounter singularities. Therefore, to warrant full agility and maneuverability of the quadrotor, it is required to have a quadrotor model that avoids such singularities.

 

In our previous research, we developed a quaternion-based model for certain aggressive maneuvers without singularities. These maneuvers include high speed vertical power looping and helix power looping. Both the controlled loop following (CLPF) method and the full quaternion differential flatness (FQDF) method are implemented on the quadrotor to enable its capability of aggressive maneuvers. With quaternion-based modeling, the quadrotor can accomplish those designed aggressive trajectory without singularity in the rotational space.

 

Although quaternion-based models can eliminate the singularity in rotation (due to the Euler sequence), there still exist other types of singularities for maneuvers in head-direction and thrust-direction. A UAV usually equips with camera or other type of sensing device, which is mounted on the front or top of the UAV and which points towards the forward direction. Keeping those sensing devices pointing at the desired orientation can be a difficult task because the definition of reference heading direction vector introduces singularity. Such singularity has been demonstrated explicitly in mathematical modeling. As shown in our 2020 IMECE paper, when a quadrotor’s thrust vector aligns with the plane of heading direction, the feedback error of the yaw controller for the quadrotor increases abruptly and becomes discontinuous. This can render the mixer of the quadrotor saturated.

 

Traditionally, when a quadrotor mixer sees saturation, it will implement one of two strategies: (i) it either ignores the saturation or (ii) it off-sets thrust to compensate insufficient torque. The first strategy, ignoring the saturation, can help translational tracking but it inevitably causes a large error in the heading direction (sensor direction). The second strategy, offsetting the thrust, can provide enough head room for the torque in the direction of the normal to the plane of the motors but it can deviate the quadrotor from the desired trajectory. Either strategy may result in poor performance of the quadrotor in tracking.

 

In this article, a special strategy in dealing with the afore-mentioned singularity and saturation is proposed. In this strategy, pre-yawing is adopted when a quadrotor’s current heading direction is approaching singularity. As shall be shown, by pre-yawing before the quadrotor passes through the singularity, it can converge toward the heading direction faster after exiting the singularity. Also, the pre-yawing strategy can reduce or avoid mixture saturation by reducing the feedback error. Therefore, the proposed strategy can also reduce the deviation from the desired trajectory. The benefits of this new control strategy in tracking are demonstrated through comparison with some existing quadrotor models in the literature.

Presenting Author: Haowen Liu University of Southern California

Authors:

Haowen Liu University of Southern California
Bingen Yang University of Southern California