Parametric and Implicit Features-Based UAV–UGVs Time-Varying Formation Tracking: Dynamic Approach

Flexible and robust Time-Varying Formation (TVF) tracking of Unmanned Ground Vehicles (UGVs) guided by an Unmanned Aerial Vehicle (UAV) is considered in this paper. The UAV–UGVs system control model is based on leader-follower approach, where the control scheme consists of two consecutive tasks, namely, deployment task and TVF tracking. Accordingly, two novel nonlinear controllers are proposed for controlling the UGVs formation. First, unlike the classical frameworks on UGVs formation tracking, for which only particular shapes are handled (e.g. circle, square, ellipse), we propose a UGVs deployment-controller ensuring to reach free-formation shapes. The key feature is in using the estimated implicit representation of the desired formation shape as a potential function to generate the UGVs reference trajectory. Second, in the TVF tracking task, a robust cascaded velocity/torque controller for UGVs is proposed based on kinematic and dynamic models. Differently from the classical backstepping framework, the key idea is in introducing an auxiliary control input, in such a way that the overall UGV dynamics is converted into a simpler and modular control structure. As such, the auxiliary input is used to control indirectly the actual UGVs velocity vector. A signum term is added to the torque-input to compensate for the unknown external disturbances and unmodeled dynamics. Numerical simulation shows the effectiveness of the proposed formation controllers compared with the case when the perfect velocity-tracking assumption holds. Experimental results are further provided using three festos Robtino robots to show the validity of the proposed TVF tracking velocity-control scheme.

FAST RRT* 3D-Sliced Planner for Autonomous Exploration Using MAVs

This paper addresses the challenge to build an autonomous exploration system using Micro-Aerial Vehicles (MAVs). MAVs are capable of flying autonomously, generating collision-free paths to navigate in unknown areas and also reconstructing the environment at which they are deployed. One of the contributions of our system is the “3D-Sliced Planner” for exploration. The main innovation is the low computational resources needed. This is because Optimal-Frontier-Points (OFP) to explore are computed in 2D slices of the 3D environment using a global Rapidly-exploring Random Tree (RRT) frontier detector. Then, the MAV can plan path routes to these points to explore the surroundings with our new proposed local “FAST RRT* Planner” that uses a tree reconnection algorithm based on cost, and a collision checking algorithm based on Signed Distance Field (SDF). The results show the proposed explorer takes 43.95% less time to compute exploration points and paths when compared with the State-of-the-Art represented by the Receding Horizon Next Best View Planner (RH-NBVP) in Gazebo simulations.

COAA* — An Optimized Obstacle Avoidance and Navigational Algorithm for UAVs Operating in Partially Observable 2D Environments

Obstacle avoidance and navigation (OAN) algorithms typically employ offline or online methods. The former is fast but requires knowledge of a global map, while the latter is usually more computationally heavy in explicit solution methods, or is lacking in configurability in the form of artificial intelligence (AI) enabled agents. In order for OAN algorithms to be brought to mass produced robots, more specifically for multirotor unmanned aerial vehicles (UAVs), the computational requirement of these algorithms must be brought low enough such that its computation can be done entirely onboard a companion computer, while being flexible enough to function without a prior map, as is the case of most real life scenarios. In this paper, a highly configurable algorithm, dubbed Closest Obstacle Avoidance and A* (COAA*), that is lightweight enough to run on the companion computer of the UAV is proposed. This algorithm frees up from the conventional drawbacks of offline and online OAN algorithms, while having guaranteed convergence to a global minimum. The algorithms have been successfully implemented on the Heavy Lift Experimental (HLX) UAV of the Autonomous Robots Research Cluster in Taylor’s University, and the simulated results match the real results sufficiently to show that the algorithm has potential for widespread implementation.