@misc{9246959,
  abstract     = {{In the past, robots were goal-based agents and experienced a more limited world. These robots needed to be instructed on each individual step or action to complete a larger task. They worked in static and well-modelled environments and run predefined behaviours where their missions consisted of repeating a single task (e.g. assembling and vacuum cleaning). Nowadays the world of robots and autonomous vehicles is growing continuously making our environment more robotised due to advanced research and development in electronics (storage and computation power) and Artificial Intelligence (AI) algorithms.
 In fact, by fully exploiting the robot capabilities and taking advantage of the AI algorithms we are changing our well-being and productivity in different sectors, particularly in industry, health-care and public safety. These technological advances opened up the opportunity for new applications. For instance, we have seen an increasing number of practical applications using multi-robot systems (or robot fleet) within industry, search and rescue (SAR), space exploration and other areas, to address specific mission scenarios where the control techniques are based on the field of multi-agent networked control in robotic networks. In particular, the powerful concept UAV-UGV cooperation where unmanned aerial vehicles (UAVs) and unmanned ground vehicles (UGVs) collaborate to create a more robust, efficient, and versatile solution for complex missions like disaster response, agriculture, or defence.
 In this degree project, we present two distinct problems in the engineering fields of Automatic Control and AI Robotics. In the first part, we investigate how to solve a Rendezvous landing problem between a quadcopter (UAV) and a mobile robot (UGV). Next, in the second part we deal with the concept of intelligent behaviour that requires human-level cognitive capabilities of which a central problem is selecting the action to do next or the so-called deliberative acting. Thus, to enable such sophisticated deliberative skills, we investigate how to design the robot to properly solve an AI planning problem (also known as Task Planning) in Robotic Inspection domain. These two problems can be related and would complement each other in certain inspection missions, for example in construction/industrial scenarios, deploying such aerial and ground vehicles.
 To this end, the goal of the project first part is to design an autopilot that can execute a Rendezvous landing algorithm, which is an important problem in autonomous flight. The main steps of the algorithm consist of: tracking, alignment, descent, retardation (shutting of propeller motors) and touch-down. Specifically, the autopilot’s task is to force the quadrotor UAV "DJI Matrice 100" (a common research platform) to track a quadruped robot "Spot" (from Boston Dynamics) and subsequently land on it.
 In this study, the dynamics of both quadrotor and quadruped systems were examined and presented, where the models are relatively simplified or necessarily approximated as first or second-order systems, in similar fashion as in various literature. Further, the Rendezvous algorithm uses Robot Operating System (ROS) to command and control the flight of the quadrotor UAV and the movement of the mobile ground robot. Substantially, the algorithm makes use of the ROS interface (including services, subscribed topics, and published topics) and runs a fourdimensional PID position-tracking controller implemented in Python to control the four basic movements of the quadrotor: thrust, roll, pitch and yaw. The objective of this controller is to compute and send correction commands to guide the UAV drone to track the target robot platform on the ground, as well as to accomplish the alignment of the acting unmanned vehicles in the horizontal plane and the landing operation at the end.
 Simulated experiments are conducted using the ROS-based UAV simulator provided by WARA-PS framework (WASP Research Arena Public Safety), as well as using a two-wheeled simulated robot (TWR), which is a newly constructed URDFbased simulation model for simulating the movement of Spot robot in the rendezvous manoeuvre. The obtained results confirm the satisfactory performance of the proposed algorithm (implemented in Python) by illustrating the convergence graph using the PID position-tracking controller.
 Further, for reinforcing the performance of this Rendezvous algorithm we additionally investigate other robust control strategies using the promising Model Predictive Control (MPC) approach, where theoretical development for three different advanced MPC-based methods is presented in this report, in which interaction between platforms can also be included in cooperative manner. However, the practical implementations of this second approach using MPC, including simulations and real-world flight tests and experiments, will be addressed in future work.
 In the second work, the goal is to formulate and implement a planning model in Planning Domain Definition Language (PDDL) that can solve an inspection planning problem by generating a sequence of composite (abstract) actions. That is, as a result we obtain a high-level goal plan to perform certain tasks and achieve the desired objectives (like "spot-inspect" or "uav-inspect"). In fact, one possible solution plan to a small simulated inspection mission example is successfully obtained using our implemented PDDL-model, composed of domain and problem files, as input to a suitable off-the-shelf AI-based solver (task planner). Next, this generated highlevel PDDL solution plan is expanded (refined) into executable Task Specification Tree (TST), using TST Factory inWARA-PS, to obtain at the end a possible cooperative goal plan consisting of detailed elementary (concrete) actions, like "move-to" and "fly-to". Testing the applicability of the resulting PDDL plan of actions in realworld problems will be performed in future research work.}},
  author       = {{Mohamad, Hicham}},
  language     = {{eng}},
  note         = {{Student Paper}},
  title        = {{UAV-UGV Rendezvous and AI Planning for Robotic Industrial Inspection Based on ROS and WARA-PS}},
  year         = {{2026}},
}