Anders Lager's PhD Defense

2025/09/02

Anders Lager, ARRAY Phd student will defend his PhD thesis on 24th October 2025 at 09:15 in Västerås Campus

Title:

Task Planning of Industrial Mobile Robots in Collaborative Dynamic Environments

Date and time: October 24th, 2025 09:15

Room: Room My and Teams Meeting

Committee:

Sven Koeing, University of California

Milica Petrovic, University of Belgrade

Luca Bascetta, Politecnico di Milano

Nng Xiong, Mälardalen University (reserve)

Advisors:

Alessandro Papadopoulos, MDU

Thomas Nolte, MDU

Branko Miloradovic, MDU

Giacomo Spampinato, ABB Robotic

Abstract:

Over the past decades, industrial robotics has transitioned from fixed, single-purpose machines to flexible, collaborative mobile systems capable of navigating complex factory environments. Today’s manufacturing demands, driven by labor scarcity, the need for rapid reconfiguration, and advances in AI and sensing, require robots to perform increasingly sophisticated, non-repetitive tasks alongside human workers. Designing and executing efficient multi-robot missions in such dynamic, human-in-the-loop settings presents multiple challenges: expressing high-level production requirements in a planner-friendly way, handling unexpected execution errors, scaling to large task allocations, and accounting for uncertainties in task durations and human behavior.
This thesis introduces an intuitive task modeling formalism and a suite of algorithmic methods that address these challenges end-to-end. First, we propose a domain-expert-friendly syntax for defining single-robot production missions, automatically generating problem definitions compatible with diverse off-the-shelf planners. To support rapid recovery from errors, we present task roadmaps, a novel planning algorithm that reuses the original search tree to accelerate replanning when execution deviates. We extend the formalism to a multi-robot kitting use case with alternative task locations and introduce a scalable, clustering-based approach to maintain computational tractability.
Recognizing the inherent uncertainties of human-robot collaboration, we further develop a collaborative stochastic task planning framework that integrates human risk preferences and models variability in task and routing durations. Finally, we tackle a collaborative production scenario with complex cross-schedule dependencies, proposing a stochastic scheduling method that generates optimized, deadlock-free plans while balancing efficiency with human well-being.
Extensive simulations and experiments grounded in real-world applications demonstrate that our methods significantly improve planning efficiency, robustness, and adaptability in dynamic industrial settings, paving the way toward more resilient, human-centric robotic automation.