The thesis investigates trajectory tracking over lossy communication networks using model predictive control, with a focus on mobile robots that rely on remote computation via edge or cloud resources. In such systems, networkinduced issues, such as packet loss and delays, can severely degrade control performance. Addressing these challenges is critical for ensuring safety and efficiency in industrial automation and robotics, especially with the increasing adoption of wireless communication technologies like 5G and 6G. The core objective is to develop a robust control strategy that allows accurate tracking of general time-varying trajectories despite network imperfections. The thesis introduces a time-varying artificial reference model predictive control approach and compares two estimation strategies: an optimistic approach that assumes received inputs were correctly applied and a novel pessimistic approach that assumes the opposite. The strategies are analysed in terms of avoidance of steady-state inputs and control performance under stochastic packet loss modelled by Bernoulli distributions. In addition, a general framework to analyse different estimation strategies is developed. Through extensive simulations, including scenarios with double integrator dynamics and an inverted pendulum on a cart, the pessimistic approach combined with the proposed time-varying model predictive control continually outperformed the optimistic strategy, particularly in high-loss network conditions. The results show a reduced mean square error and an increased probability of using optimal control inputs, enabling more reliable trajectory tracking. The study also introduces decision tools based on network parameters and probability of consistencies to guide the selection of appropriate time horizons and estimation strategies. The thesis contributes a practical simulation-verified framework for reliable control in lossy network environments and provides insights applicable to both academic research and industrial deployment. Furthermore, by optimizing control performance and reducing unnecessary communication and computation, the proposed methods support sustainable development goals related to innovation, infrastructure, and responsible consumption.
Axel Abrahamsson (Wed,) studied this question.