Pure Inertial Navigation is practically unfeasible with low cost sensors; usually aiding sources like GPS (in terrestrial/airborne applications) or acoustic sensors (in marine applications) are fused via an Extended Kalman Filter to provide non drifting estimates of the navigation outputs (position, velocity, attitude). When this types of aids are not available, artificial vision can be used to provide the navigation filter with position and velocity fixes. Within this context, the non-white noise spectrum and the presence of outliers in the vision-based measurements pose serious convergence problems. Appropriate estimation and filtering techniques must be developed which possess enough robustness to the class of measurement errors present in the vision data.

Research partially supported by Northrop Grumman Italia.

Accurate Navigation is necessary for work class Remotely Operated Vehicles (ROVs) operating up to thousands of meters of depth. Inertial Navigation and sensor fusion with all the available measurements (USBL, DVL, Depth sensors etc.) is a key technology. Navigation Filters must deal with sensor delays, sensor black-outs, discontinuous sampling, time-varying biases and drifts and non-white noise.

Research funded by Saipem-Sonsub.

Autonomous Formation flight and aerial refueling are currently two important aspects in aerospace research.  In autonomous formation flight, precise and stable absolute and/or relative positioning of each single vehicle is necessary. Aerial refueling can be performed with two methods currently: the Boeing flying boom (with issues very similar to formation flight) and the probe-and-drogue method. In the latter approach, a "soft basket" attached to a flexible hose floats behind the tanker and its position is strongly affected by the wake of the tanker, by the wake of the approaching UAV and by the environmental conditions. Although a GPS system could be used, in the presence of potential disturbances, or loss or intentional denial of GPS signal, a potentially interesting alternative is the use of artificial vision for measuring relative displacement between the aircraft or with the drogue. 

[Relative position and attitude estimation using a simulated formation flight testbed using an IR camera, and a robot for moving a scale model aircraft.]

Research developed in cooperation with West Virginia University.

Recently Northrop Grumman Italia financed a PhD Curriculum under the supervision of Lorenzo Pollini.

Conventional waypoint guidance techniques do not allow specifying the desired waypoint approach path, while this feature may be usefull in many practical cases. The proposed approach exploits Fuzzy Logic as a mean of easy specification of the desired trajectories in the vicinity of the waypoint.  Closed loop stability and performance of such a system needs to be studied and assessed.

The aim of this research activity is to study and TO prototype novel real-time obstacle detection system based on stereoscopic vision. The system is designed and tested with an Unmanned Ground Vehicle (UGV) operating in an unstructured environment.  This research investigates the possibility of fusing together feature-based (SIFT, SURF) and depth-map based approaches;  Algorithms are implemented in real-time using C++ and NVIDIA-CUDA. [ Sample Movie ]

Research partially supported by NVidia.

The Null Space Based Behavioral Control (NSBBC) algorithm is a Control Architecture which relies on simple controllers/behaviors, coordinated by a supervisor, to achieve simultaneous goals like reaching a target while avoiding obstacles. The supervisor assigns priorities to each behavior dynamically, and is primarily responsible of achieving or not all the control objectives. Universal rules for designing the supervisor do not exist; thus this research activity aims at finding a set of rules which may help the developer of the supervisor in building and verifying if a set of switching rules can drive the vehicle to the goal avoiding obstacles, in a stable fashion.  [ Sample Movie 1]  [ Sample Movie 2]

Decentralized control of a large number of agents acting as a single swarm is an active field of research, but the problem of how to instruct a swarm of autonomous vehicles to perform certain tasks, and in particular, how to interface the swarm to a human operator has not been very well studied yet. The aim of this research activity is to study and to prototype a Human Swarm Interface (HSI) which allows the swarm operator to intuitively control and monitor a swarm position and shape over time, to asses the degree of completion of the assigned task visually on a map of the scenario, and to actively control the swarm in closed loop by adapting quickly his/her commands to the current swarm state and unforeseen environmental changes. Within this approach, the swarm position and shape is defined using an Abstraction based approach where each single swarm agent estimates the swarm state via a distributed consensus algorithm and moves according to a steepest descent algorithm toward minimization of the error between human operator command and swarm state.  [ Sample Movie - Why everybody needs obstacle avoidance]

This research activity studies a novel approach for deployment of a swarm of heterogeneous autonomous vehicles. Each vehicle is treated as the agent of a network, which cooperates in order to cover a given area according to its own capabilities. A general framework is sought which aims at providing tools for solving a large class of coordination problems. The capabilities of each agent are modeled with Descriptor Functions; the sum of these functions constitutes the swarm’s descriptor. The goal of the swarm is to match a desired descriptor, by minimizing an appropriate cost functional. 

State-Dependent Riccati Equation (SDRE) control method is a somewhat recent control technique; one of the main advantages of this method is that it is systematic, and it can be applied to a large class of nonlinear systems. Unfortunately the stability properties of SDRE-controlled systems can be guaranteed only in a small neighborhood of the origin, so there is no guarantee of global stability in the general case. This research activity investigates application and stability issues of SDRE control. 

The Takagi–Sugeno (TS) fuzzy model theory proven to be useful in the description of nonlinear dynamic systems as a mean for the blending of models obtained by local analysis. Such descriptions are referred to as model-based fuzzy systems (MBFS). In addition, the TS approach can be used for the synthesis of fuzzy gain-scheduled controllers. Several issues arise in this approach: scheduling grid points selection, shape of interpolating functions, stability in-between scheduled controllers, off-equilibrium linearization etc. The Linear Matrix Inequalities (LMI) tool could  be used to cast several of those issues into a unique framework for stability and performance. 

Power generation from biomass has emerged as a very interesting complement to conventional sources of energy because of its contribution to the reduction of the greenhouse effect. Among the alternatives to produce electricity from biomass, gasification turns out to be cheaper while having higher efficiency, though very low. Additionally, downdraft Gasifiers with throat are known to produce the best quality gas for burning in internal combustion engines. As the installation can process different types of biomass, the control system needs to adapt to the type of biomass, and also its changing conditions: input temperature,  moisture,  etc in order to keep efficiency high and provide the engine with a good quality of syngas.

Research funded by Regione Toscana.

A Dynamic Positioning (DP) system can be defined as a ship control system where the ship position is controlled exclusively by means of the thrusters and the main propellers of the ship, under the disturbance loads due to wind, waves, and currents which change over time both in spectral signature, direction and magnitude. Since, only position and attitude sensors are available for most commercial ships, usually a model-based output feedback control is used where the velocities and disturbances are estimated by using a nonlinear state observer. An alternative possibility is to measure the vessel acceleration and use it to better estimate the higher frequency disturbances in order to suppress them.

Research Funded by Saipem-Sonsub.

This research activity investigates novel ideas the field of aircraft pilot visual aiding using perspective flight path display and guidance tunnel. Possible applications include: precision atmospheric survey, precision farming, generic guidance and navigation aid, terminal flight conditions. The design requirements and points of novelty which are driving this research are: low cost, low weight and low consumption, independence from aircraft dynamics, capability to operate independently from the presence or not of other avionics, independence from air data, that is, fully inertial and GPS enabled; finally the system should improve the capability to recover when the path following error is very large, and the pilot’s situation awareness in terms of the future effects of his current flight commands. Early testing of the system started in 2009 yielding very interesting results. 

Research developed in cooperation with University of Bologna, Forli.

The autopilot market for small and research UAVs offers several products; these systems are viable and accessible solutions for making a UAV fly but, although claimed widely configurable and customizable are not sufficiently open as platforms for implementation and development of custom guidance, navigation and control (GNC) systems. The ICARO project aims at providing the small UAV community with a valid alternative to other commercial products. Main features of the ICARO autopilot are: double CPU for computing redundancy, full suite of hardware input interfaces, specially suited for rapid-prototyping of GNC and fault-tolerance algorithms, completely open and programmable. [ Sample Movie - ICARO mounted on a quadrotor UAV]

Grappa is a fragrant grape-based pomace brandy of between 37.5% and 60% alcohol by volume. During red wine vinification, for instance, the pomace is left to soak in the must for the entire fermentation period and is thus fermented. After fermentation is complete the pomace wine is distilled. Large companies uses mechanized systems to feed pomace to the distiller tank. Often extraneous objects like iron wires, lost scissors, hammers, ropes etc may be present in pomace loads and must be removed prior to entering the distiller in order to avoid mechanical damages or hitches. A vision system with a robotic grabber was designed in order to detect, classify and remove from the feeding belt anything but pomace.  The system was designed by Eontych for Frilli Impianti, with the collaboration of University of Pisa for the motion control. [ A sample movie shot during integration testing]

Research Funded by Eontych.

The main goal of ENFICA-FC programme is to study and demonstrate the possibility of obtaining an all electric aircraft through the integration of fuel-cells technology as main power-supply system. During the three years research activity, the consortium designed, developed and installed a fuel-cell based power system in a ultra-light aircraft with a brushless electric engine driven by a specifically designed lightweight power inverter. A new avionics with glass cockpit was designed and implemented for supervision and control of the complete power chain. The aircraft is currently awaiting approval by the Italian Aviation Administration (ENAC) in order to start airborne testing.  

Research Funded by EC - 6th FP

A picture of me onboard.

Enfica-FC project website: www.enfica-fc.polito.it

This research activity investigates  a novel approach to motion platform design for immersive motion simulators which employs an anthropomorphic industrial robot for motion of the cabin in place of the widely used Stewart’s platform. The whole processing chain from vehicle simulation, to washout filters and platform Inverse Kinematic (IK) needs to be redesigned in order to take advantage of this new motion system.  Several experiments have already  been carried out using a modified KUKA Robocoaster and new washout and IK algorithms.   [ Sample Movie - Car Driving ]

Research developed in cooperation with Max Planck Institute – Tubingen.

Computer Aided Manufacturing (CAM) software is usually used to program Computer Controlled Machining Centers. Industrial anthropomorphic robots can be used in place of CNC machines where very large pieces must be produced. Since anthropomorphic robots have more degrees of freedom than those needed by standard processes (like milling, painting, polishing etc) but a more complex kinematics with respect to crane-like 5-D CNC machines, and the robotic workcells often contain turn tables, linear 1-DoF or 2-DoF slides, specific optimization algorithms are needed in order to manage the additional degress of freedom, exploit the full working envelope, avoid singularities and maintain a high robot stiffness. 

[Photos and video snapshots]

Research Funded by QDesign.

Currently Haptic augmentation systems for remotely operated vehicles (RPVs) are designed in order to help directly the pilot in his/her task by pulling the stick in the correct direction for the achievement of the task, or by changing stick stiffness in order to facilitate or oppose to certain pilot’s actions. The sense of touch could be used, as originally intended, to provide the pilot with an additional source of information that would help him indirectly instead. Thus this research aims at designing novel haptic augmentation schemes which increase the situation awareness, that is infer a better knowledge of system status and of its external disturbances. This implies that the haptic feedback must trigger the pilot prior knowledge of the force response/dynamics of the vehicle he is piloting; as a consequence, the impact of pilot training with a  specific force feedback must be accurately understood.

Research developed in cooperation with Max Planck Institute – Tubingen.