In this section there are some bites of my current and past activities about AGV (Automatic Guided Vehicle) both on pure research and industrial fields.
Hereafter a sequence of the results obtained in end august 2011 within the AGILE project for which we developed an automatic system for pallets identification, localization and picking. The sequence shown were obtained in an industrial plant of our partners Digipack, Sanovo group, of Galliera Veneta (PD).
This kind of flexibility we believe will revolutionize logistics: poorly structured warehouses, corridors, external truck unloading stations will be incorporated in plant automation.
|AGILE: Intelligent AGVs for autonomous pallet Identification/Localization/Transportation in non-structured industrial Environments
Nowadays Automated Guided Vehicles (AGVs) are used in many production environments and warehouses such as the automotive industry, logistics or container harbours. In general, the AGVs are highly expensive and often not flexible. The path of an AGV is in most cases strictly defined and cannot be changed on demand. The environment, in which the AGVs are working must therefore be well structured and paths must show a high repeatability rate. Many reconstructions in the infrastructure are needed to install and make the system work. Due to high costs and their limited robustness and flexibility AGVs are not often used in medium sized enterprises and especially not in non-structured environments.
Therefore the idea of the AGILE Eurostars project is to develop a small, robust and flexible AGV for semi- or non-structured warehouses targeted to achieve the following:
The AGILE system is composed of a Real Time processor, a safety laser and a camera.
In the pictures, from top-down:
The AGILE project partners are the SME Fusion Systems Gmbh Germany and Digipack Italy. The sole academic partner is the Department of Mechanical and Structural Engineering of the University of Trento.
|Path generation and control is the problem of determining a feasible set of commands that will permit a vehicle to move from an initial state to a final state following a desired geometrical figure in space while correcting for deviations in real time.
While this problem can be solved for manipulators by means of inverting nonlinear kinematics, the common inverse problem for mobile robots is that of inverting nonlinear differential equations. A basic method is therefore to plan a geometric path in the surface of motion, generally a 2D space, and conceive a suitable control strategy to force the vehicle to follow it. If the path is feasible its tracking will be accurate, otherwise there will be non negligible errors in the executed path. If one plan a continuous curvature path, than can be sure of its compatibility with respect to kinematics and partially to dynamics if the maximum rate of curvature variation is taken into consideration. This for a huge variety of vehicles. As a matter of fact differential drive, car-like and all wheel steering vehicles have constraints in curvature variation while moving.From those considerations we started to develop a Polynomial Curvature Sliding control, PC-Sliding [C37], a novel RT procedure for planning and control that can be summarised as follows. The steering commands are designed by means of the polynomial curvature model applying a two-point boundary value problem driven by the differential posture (pose plus curvature). While following the path the vehicle re-plans iteratively the path with a repetition rate that must not necessarily be deterministic. To the actual curvilinear coordinate it is added a piece forward, than computed the corresponding posture in the original planned path, finally replanned the differential path steering the vehicle from the actual posture to the one just computed. The result is to force the vehicle to correct for deviations while sliding over the desired path.
Advantages of the proposed method are its essentiality thanks to the use of the same strategy both for planning and control. Controlling vehicles in curvature assures compatibility with respect to kinematics and partially to dynamics if the maximum rate of curvature variation is taken into consideration. The method doesn’t need chained form transformations and therefore is suitable also for systems that cannot be transformable like for example non-zero hinged trailers vehicles (Lucibello 2001). Controls are searched over a set of admissible trajectories resulting in corrections that are compatible with kinematics and dynamics, thus more robustness and accuracy in path following.
|Starting from the prototype developed for ItalProject and with the collaboration of Digipack, an industrial transpallet was optimized and made it autonomous.Main characteristics of the industrial system are: obstacle avoidance capability, on-line calibration, more than 30 meters of blind path that means without artificial references, continuous curvature path planning.
The industrial research was in collaboration and of benefit to ItalProject, http://www.italproject.net/main_it.asp, and Digipack, http://www.digipack.it/
Since market introduction, year 2006, the new vehicle has gained a good reputation and market share in the field of automatic warehouses.
|Starting from a manual transpallet it was developed an Autonomous Guided Vehicle AGV.
Main characteristics are: obstacle avoidance, on-line calibration, continuous curvature path planning, more than 30 meters of blind path that means without artificial references.The industrial research was in collaboration and of benefit to ItalProject (PD),
|Development of a path tracking algorithm on a straight line for a double drive wheel vehicle for the company Fereng.
It was designed and implemented an original algorithm for the vehicle that has two driver-steering wheel, therefore 4 degrees of control and 3 dof by means of a instantaneous centre of rotation control. The company had previously implement an algorithm capable of subjecting the vehicle to the line in about 15 meters, the algorithm developed by us allowed the alignement in less than 2 meters.
|This is a differential drive prototype AGV designed and assembled during my previous position at the CISAS lab.This prototype has the aim to be a test bed for navigation and control algorithms.
The system has two independent motors, an RT control algorithm, two encoders, a gyro and a safety laser scanner to estimate the presence of obstacles and acquire a local map of the environment.
The research was focused principally toward the following topics: