Until recently, the universally adopted paradigm for industrial robotics entailed the robot confinement in a dedicated operating space for safety issues. In the last few years, a new concept of manipulator is arising: the so-called collaborative robot (cobot). This represents a true fellow of the human operator since it works alongside him/her, sharing the same workspace. To ensure a safe human-robot cooperation (HRC), cobots have been equipped with safety functionalities such as Speed and Separation Monitoring (SSM) and Power and Force Limiting (PFL). These safety measures prescribe a reduction of the robot speed proportionally to the proximity to the human operator and require the cobot to work with limited energy in the vicinity of the human. However, the intervention of these criteria, might penalize the productivity of the human-robot team. This thesis presents a real-time trajectory optimization method for a collaborative robot that simultaneously minimizes the risk of collision with human as well as the robot motion duration. The proposed optimization method is based on a genetic algorithm which randomly modifies the parameters of nominal trajectory of the robot and selects the one optimizing the trade-off between shorter cycle-time and the proximity to the human operator. Each trajectory is simulated on a digital twin of the collaborative workspace, which produces a perfect replica of the robot motion and of the workspace occupancy. Indeed, a smart 3D vision device continuously collects data about the work-cell volume occupied by the human operator in the work-cell and produces a probabilistic occupancy 3D grid, which is used to evaluate the long-term human occupancy. This method is validated in a realistic industrial working scenario, where the ABB dual-arm collaborative robot (YuMi) and a human operator work together to perform an assembly task. Experimental results proved that, by applying the proposed method, the robot adapts its trajectory based on the specific occupancy of the human fellow and it is effective in achieving the best trade-off between productivity and safety.
Fino al recente passato il paradigma universalmente adottato nell’ambito della robotica industriale prevedeva il confinamento dei robot in ambienti protetti. Negli ultimi anni, sta emergendo un nuovo concetto di manipolatore: il robot collaborativo (cobot). Esso si configura come un vero e proprio partner dell’operatore umano in quanto lavora insieme a lui/lei, condividendo il medesimo spazio di lavoro. Per garantire una sicura cooperazione uomo-robot (HRC), i cobot sono stati dotati di funzionalità di sicurezza, quali il monitoraggio di velocità e distanza e la limitazione di potenza e forza. Queste funzionalità di sicurezza richiedono la riduzione della velocità del robot proporzionalmente alla sua vicinanza all’operatore e vincolano il cobot a lavorare con un’energia limitata in prossimità all’essere umano. Tuttavia, l’applicazione di questi criteri di sicurezza comporta una penalizzazione della produttività del team uomo-robot. Questa tesi propone un metodo di ottimizzazione real-time della traiettoria per un robot collaborativo, che minimizza simultaneamente il rischio di collisione con l’uomo e la durata del movimento. Il metodo di ottimizzazione proposto si basa su un algoritmo genetico che modifica casualmente i parametri della traiettoria nominale del robot e seleziona quella che ottimizza il trade-off tra minor tempo ciclo e prossimità all’operatore umano. Ogni traiettoria è simulata su un digital twin, che produce una fedele replica del movimento del robot e dell’occupazione dello spazio di lavoro. I dati sul volume della cella di lavoro occupato dall’operatore sono raccolti da un sistema smart di visione 3D, che produce una mappa volumetrica probabilistica, utile a valutare l’occupazione umana a lungo termine. Questo metodo è stato validato in uno scenario di lavoro industriale, dove il robot collaborativo a due bracci di ABB (YuMi) e un operatore umano collaborano per l’esecuzione di un task di assemblaggio. I risultati sperimentali hanno dimostrato che, applicando il metodo proposto, il robot adatta la sua traiettoria in base all’occupazione del partner umano, raggiungendo così il compromesso tra produttività e sicurezza.
Robot trajectory optimization for long-term human occupancy in collaborative frameworks
Cristantielli, Davide
2019/2020
Abstract
Until recently, the universally adopted paradigm for industrial robotics entailed the robot confinement in a dedicated operating space for safety issues. In the last few years, a new concept of manipulator is arising: the so-called collaborative robot (cobot). This represents a true fellow of the human operator since it works alongside him/her, sharing the same workspace. To ensure a safe human-robot cooperation (HRC), cobots have been equipped with safety functionalities such as Speed and Separation Monitoring (SSM) and Power and Force Limiting (PFL). These safety measures prescribe a reduction of the robot speed proportionally to the proximity to the human operator and require the cobot to work with limited energy in the vicinity of the human. However, the intervention of these criteria, might penalize the productivity of the human-robot team. This thesis presents a real-time trajectory optimization method for a collaborative robot that simultaneously minimizes the risk of collision with human as well as the robot motion duration. The proposed optimization method is based on a genetic algorithm which randomly modifies the parameters of nominal trajectory of the robot and selects the one optimizing the trade-off between shorter cycle-time and the proximity to the human operator. Each trajectory is simulated on a digital twin of the collaborative workspace, which produces a perfect replica of the robot motion and of the workspace occupancy. Indeed, a smart 3D vision device continuously collects data about the work-cell volume occupied by the human operator in the work-cell and produces a probabilistic occupancy 3D grid, which is used to evaluate the long-term human occupancy. This method is validated in a realistic industrial working scenario, where the ABB dual-arm collaborative robot (YuMi) and a human operator work together to perform an assembly task. Experimental results proved that, by applying the proposed method, the robot adapts its trajectory based on the specific occupancy of the human fellow and it is effective in achieving the best trade-off between productivity and safety.| File | Dimensione | Formato | |
|---|---|---|---|
|
2021_04_Cristantielli.pdf
solo utenti autorizzati dal 07/04/2022
Dimensione
16.03 MB
Formato
Adobe PDF
|
16.03 MB | Adobe PDF | Visualizza/Apri |
I documenti in POLITesi sono protetti da copyright e tutti i diritti sono riservati, salvo diversa indicazione.
https://hdl.handle.net/10589/175581