In this paper, platoons of autonomous vehicles operating in urban road networks are considered. From a methodological point of view, the problem of interest consists of formally characterizing vehicle state trajectory tubes by means of routing decisions complying with traffic congestion criteria. To this end, a novel distributed control architecture is conceived by taking advantage of two methodologies: deep reinforcement learning and model predictive control. On one hand, the routing decisions are obtained by using a distributed reinforcement learning algorithm that exploits available traffic data at each road junction. On the other hand, a bank of model predictive controllers is in charge of computing the more adequate control action for each involved vehicle. Such tasks are here combined into a single framework: the deep reinforcement learning output (action) is translated into a set-point to be tracked by the model predictive controller; conversely, the current vehicle position, resulting from the application of the control move, is exploited by the deep reinforcement learning unit for improving its reliability. The main novelty of the proposed solution lies in its hybrid nature: on one hand it fully exploits deep reinforcement learning capabilities for decision-making purposes; on the other hand, time-varying hard constraints are always satisfied during the dynamical platoon evolution imposed by the computed routing decisions. To efficiently evaluate the performance of the proposed control architecture, a co-design procedure, involving the SUMO and MATLAB platforms, is implemented so that complex operating environments can be used, and the information coming from road maps (links, junctions, obstacles, semaphores, etc.) and vehicle state trajectories can be shared and exchanged. Finally by considering as operating scenario a real entire city block and a platoon of eleven vehicles described by double-integrator models, several simulations have been performed with the aim to put in light the main features of the proposed approach. Moreover, it is important to underline that in different operating scenarios the proposed reinforcement learning scheme is capable of significantly reducing traffic congestion phenomena when compared with well-reputed competitors.

Traditional methodologies for precise inspection of bridges (pavement, beams, column cap, column, joints and inside box girder, etc.) with By-bridge equipment, Aerial Work Platform (AWP) or via ropes have several limits that can be overcome by using Unmanned Aerial Vehicles (UAVs). The constant development in this field allows us to go beyond the manual control and the use of a single UAV. In the context of inspection rules, this research provides new inputs to the multilevel approach used today and to the methods of structural inspection with drones. Today, UAV-based inspections are limited by manual and/or semi-automatic control with many restrictions on trajectory settings, especially for areas of difficult access with Global Navigation Satellite Systems (GNSS) denied that still require the intervention of a human operator. This work proposes the use of autonomous navigation with a fleet of UAVs for infrastructural inspections. Starting from a digital twin, a solution is provided to problems such as the definition of a set of reference trajectories and the design of a position controller. A workflow to integrate a generic Bridge Management System (BMS) with this type of approach is provided.

This paper focuses on the trajectory tracking problem for a swarm of mobile agents. A kinematic model describing the interactions and evolutions of the swarm members is proposed and its main properties are analyzed emphasizing that the agents centroid is ensured to track in finite-time a given reference trajectory and that the agents reach an aggregation in finite-time in a hyper-ball moving around the centroid path. One of the main characteristics of the model is the presence of an interaction matrix, between agents coordinates, which allows to define some properties of the swarm allowing the creation of different forms of agents aggregations, i.e., spheres, ellipsoids, straight lines, etc. Indeed swarm properties related to the agents configuration around the performed path along with agents interactions and absence of collisions are analyzed depending on the chosen interaction matrix.

In this note, the reference tracking problem for teams of unmanned vehicles subject to formation constraints is solved via a model predictive control (MPC) algorithm built up in a distributed fashion. By exploiting the properties deriving from a novel kinematic description of the swarm agents, the receding horizon control (RHC) approach is properly adapted to deal with tracking and formation constraints. In particular, neighbor interactions are translated into convex conditions, thanks to an in-depth analysis of the geometric properties arising from the combined use of swarm kinematics and state predictions tubes. Experimental results on Elisa-3 robots show the applicability and effectiveness of the proposed control architecture.

In this paper, a model for a multi-agent swarm evolving in a d-dimensional space is proposed. The model consists of a balance among attraction to a given goal and attraction/repulsion among agents. One of the main characteristics of the model is the presence of an interaction matrix on agents' coordinates, defining some properties of the swarm formation. The paper focuses on the analysis of the new model highlighting the agents' steady-state behavior both in terms of static and rotating configuration around the goal. Numerical simulations are provided to support the obtained results.