Available theses (October 2011) MERLIN Group Politecnico di Milano - Dipartimento di Elettronica e Informazione
MERLIN Group 2 Luca Bascetta bascetta@elet.polimi.it Gianni Ferretti ferretti@elet.polimi.it Gianantonio Magnani magnani@elet.polimi.it Paolo Rocco rocco@elet.polimi.it Andrea Zanchettin zanchettin@elet.polimi.it + Nicola Ceriani, Amir Ghalamzan, Gianpaolo Rizzi.
MSc Theses at Politecnico di Milano 3 Expected effort Tesi 6 months full time Tesina 3 4 months full time Reviewer required yes no Maximum increment for the final grade 7/110 4/110
ROSETTA project 4
ROSETTA project 5 Rosetta Objectives Dual arm robot control - 1. Wide and flexible use 2. Easy to integrate 3. Easy to instruct 4. Easy to work with Task CAD input Knowledge base CAD production data
ROSETTA safety issues 6 Robot Co-operation Co-worker Inspector Co-existence Helper Intrusion
ROSETTA Thesis 1 7 Sensor fusion in a robotic cell for situation interpretation in the ROSETTA cell we will have both supervisory cameras and a sensor distributed along the arms how can we fuse the information coming from these sensors? can we answer questions like: there is a human in a certain position near the robot and the most probable impact point is the (head/torso/arms/legs)?
ROSETTA Thesis 2 8 Cooperation, coexistence and interference with robots depending on situation interpretation we will enforce a certain safety behaviour of the robot the main task will be suspended, or executed at lower speed exploiting redundancy to reduce danger potential, or the robot will be stopped how to select best policies for all these situations? results will be demonstrated on the FRIDA prototype
ROSETTA Thesis 3 9 Path planning with obstacle avoidance and human-like motion we have developed algorithms for collision free path planning we also have a methodology to assign human-like trajectories to a redundant robot arm can we combine these methodologies, having humanlike collision free trajectories? results will be applied to FRIDA prototype
ROSETTA Thesis 4 10 Safety-oriented control of human-like redundant manipulators methodology to classify human-like trajectories are available in the literature a sensor-based control algorithm has been recently developed can we combine the two approaches (e.g. accounting for non-human-like motion as an additional source of physiological danger)?
ROSETTA Thesis 5 11 Human-like redundancy resolution during bimanual manipulation natural exploitation of kinematic redundancy during in-hand manipulation development of a model aimed at describing the redundancy resolution of human beings during twohanded manipulation possible implementation on FRIDA
ROSETTA Thesis 6 12 Resolution of kinematic redundancy a Matlab toolbox to customize the redundancy resolution has been deployed developing an interface between this toolbox and commercial programming tools (e.g. ABB Robot Studio)
ROSETTA Thesis 7 13 Linking a safety controller to a knowledge integration framework in ROSETTA a knowledge integration framework (KIF) is being developed, where all the information concerning production and safety will be stored in the form of ontologies how can we interface a safety controller to this KIF? what are the information we want to retrieve from and store inside the server? Task Knowledge base
ROSETTA Thesis 8 14 Task-based programming of a robot within the ROSETTA project some methodologies have been developed for high-level (task based) programming of robots examples have been provided for assembly of small parts can we replicate such examples with our own prototype of FRIDA?
FIDELIO project 15 Robotized deburring of wheels
FIDELIO Thesis 1 16 Using a camera to characterize a metallic burr before deburring the wheel, burr must be localized and characterized the deburring tool correct velocity and contact force depend on the volume of the material to be removed how can the robot automatically scan a wheel localizing burrs and measuring their length/thickness?
FIDELIO Thesis 2 17 Programming an industrial robot by human demonstration we are developing a methodology for manual guidance we will have a human worker perform a first debur holding the tool at the robot endeffector several measurements will be taken what are the algorithms to be used for best replication of the human operation? demonstration on robot COMAU Smart Six
FIDELIO Thesis 3 18 Hybrid force-vision control during the robotic deburring operation the robot exploits force and vision measurements in order to track the edge vision helps in adapting to rapid curvature variation (the task execution speed can be increased) how can we detect the edge to be deburred in realtime?
Other theses in industrial robotics - Thesis 1 19 Interfacing new sensors to the robot controllers for demo purposes we have in our lab two industrial robots with open controllers soon we will have a FRIDA prototype, again with open controller there are a lot of new sensors which can be exploited for demo purposes (Kinect, Wii Mote, ) can we make FRIDA replicate our motion, captured by a Kinect?
QUADRIVIO project 20 Autonomous drive of an ATV (All Terrain Vehicle)
Quadrivio Thesis 1 21 Virtual rider of an ATV we have an instrumented ATV in our lab we want to make it partly autonomous with sensor based navigation planning the path in a rough and/or sloping terrain, taking into account the vehicle stability problem is a hard task how can the robot find its best path in a rough terrain with obstacles?
Quadrivio Thesis 2 22 ATV tracking controller we have an instrumented ATV in our lab throttle, brake and steer can be remotely actuated we want to make it partly autonomous with sensor based navigation how can we design a controller to track a desired path acting on steer, throttle and brake (and ensuring the vehicle stability)?
Quadrivio Thesis 3 23 ATV anti-rollover systems we have an instrumented ATV in our lab we have already studied the roll/tip-over problem the ATV roll/tip-over problem is extremely important for human and automatic riders as well which is the best roll/tip-over index? are we able to automatically react to an incipient roll/tip-over in order to stabilize the vehicle?
Intention estimation Thesis 1 24 Generation of human walking paths: an optimal control approach we have collected a thousand human walking trajectories using MOCAP how humans plan their paths? do they optimize an energy criterion? are we able to find the suitable cost function (of an optimal control problem) to replicate human walking paths?
25 Intention estimation Thesis 2 Estimating the human intention: a Kalman filtering approach we have already studied the intention estimation problem in the ROSETTA project (adopting a complex stochastic approach) we know how to plan human walking paths at least in simple environments walking people can be tracked by visual systems (as in the ROSETTA project) are we able to predict where human are heading to using Kalman filters based on our planner?
Continuum robots Thesis 1 26 Continuum robots: modeling and control we have studied modeling and control of flexible manipulators (space robots) there is again an increased interest in flexible robotics, in the context of continuum robots important examples of continuum robots are the tools for minimal invasive surgery how can we model and control these robots?
Robotics for kayakers Thesis 1 27 A robotic kayak ergometer for intelligent training development of a robot-based kayak training machine characterization of the paddling motion implementation of a rehabilitation-like human-robot interface utilization of the robot as an haptic interface to mimic the paddle/water interaction