Projects per year
Abstract
Autonomous aerial robots and cooperative swarms of these systems have the potential to transform our society fundamentally. They promise a world where they deliver packages and lifesaving medical supplies in minutes and efficiently perform search-and-rescue missions involving the cooperative transport of large objects in badly accessible disaster-stricken regions. However, despite significant progress in robotics over the last few decades, such systems are still far from mainstream real-world deployment. A major reason is that today’s aerial robotic systems lack guarantees of their operational safety, as they can fail dramatically in the face of uncertainty and unanticipated changes in their environment. As we expect these dynamic systems to operate with extreme agility while being close to other moving robots and humans, the lack of safety guarantees poses an immense challenge. Moreover, this challenge is magnified by the fact that aerial robotic platforms exhibit minimal resources for onboard computation, memory, communication, sensing, and actuation. Even for larger platforms with more advanced capabilities, the computational power available to implement control algorithms is typically limited in favor of running mission-dependent algorithms related to localization and sensing systems. Hence, computationally efficient, certified safe, and well-performing control algorithms that allow the navigation of aerial systems to high-level targets along prior unknown paths are of paramount importance for achieving safety-critical tasks in real-world obstacle-cluttered and dynamic environments.
This Ph.D. research has contributed to the theoretical and algorithmic development and to the numerical and experimental validation of real-time control approaches that enable cooperative aerial robotic systems to operate with guaranteed safety and various performance levels in a priori unknown and dynamic environments. The methodology is fundamentally based on the Explicit Reference Governor (ERG), which is a general framework for the closedform constrained control of pre-stabilized safety-critical nonlinear dynamical systems subject to pointwise-in-time constraints on their states and inputs. The ERG theory and its two main ingredients, i.e., the Dynamic Safety Margin
and the Navigation Field, were formally extended and specialized in dealing with dynamic aerial robot systems in scenarios requiring distributed collision avoidance and cooperative object transport capabilities. Methodological variations of the ERG based on Lyapunov level-set and invariant set theory, and receding-horizon trajectory predictions were formulated and compared in terms of performance, real-time capability, robustness, scalability, and design simplicity. Some distributed ERG formulations were proposed that rely on local information only to solve the global navigation task safely, and details for digital (i.e., discrete-time) ERG implementation were also proposed. Also, the
optimization-free ERG was demonstrated to be an efficient tool that enables broader applicability of optimization-based Nonlinear Model Predictive Control (NMPC) laws as the ERG can enforce recursive feasibility of the optimal control problem associated to an NMPC law with an arbitrarily short prediction horizon, without requiring hard-to-compute feasibility sets or terminal sets. Finally, this research has led to the first successful demonstrations of this provably safe ERG methodology on real-world aerial systems such as nine handpalm-sized nano-quadrotors and two medium-sized quadrotors, and in realistic simulators with over 30 quadrotors in the free-flight and two quadrotors cooperatively transporting a cable-suspended bar-payload.
The obtained certificates on safety, computational efficiency, goal satisfaction, robustness, performance, and scalability bring us one step closer to safely deploying highly reconfigurable, on-demand, distributed intelligent aerial robotic systems in the real world which will improve the quality of our daily lives by impacting many areas of science, technology, and society.
This Ph.D. research has contributed to the theoretical and algorithmic development and to the numerical and experimental validation of real-time control approaches that enable cooperative aerial robotic systems to operate with guaranteed safety and various performance levels in a priori unknown and dynamic environments. The methodology is fundamentally based on the Explicit Reference Governor (ERG), which is a general framework for the closedform constrained control of pre-stabilized safety-critical nonlinear dynamical systems subject to pointwise-in-time constraints on their states and inputs. The ERG theory and its two main ingredients, i.e., the Dynamic Safety Margin
and the Navigation Field, were formally extended and specialized in dealing with dynamic aerial robot systems in scenarios requiring distributed collision avoidance and cooperative object transport capabilities. Methodological variations of the ERG based on Lyapunov level-set and invariant set theory, and receding-horizon trajectory predictions were formulated and compared in terms of performance, real-time capability, robustness, scalability, and design simplicity. Some distributed ERG formulations were proposed that rely on local information only to solve the global navigation task safely, and details for digital (i.e., discrete-time) ERG implementation were also proposed. Also, the
optimization-free ERG was demonstrated to be an efficient tool that enables broader applicability of optimization-based Nonlinear Model Predictive Control (NMPC) laws as the ERG can enforce recursive feasibility of the optimal control problem associated to an NMPC law with an arbitrarily short prediction horizon, without requiring hard-to-compute feasibility sets or terminal sets. Finally, this research has led to the first successful demonstrations of this provably safe ERG methodology on real-world aerial systems such as nine handpalm-sized nano-quadrotors and two medium-sized quadrotors, and in realistic simulators with over 30 quadrotors in the free-flight and two quadrotors cooperatively transporting a cable-suspended bar-payload.
The obtained certificates on safety, computational efficiency, goal satisfaction, robustness, performance, and scalability bring us one step closer to safely deploying highly reconfigurable, on-demand, distributed intelligent aerial robotic systems in the real world which will improve the quality of our daily lives by impacting many areas of science, technology, and society.
| Original language | English |
|---|---|
| Awarding Institution |
|
| Supervisors/Advisors |
|
| Award date | 25 May 2023 |
| Publication status | Published - 2023 |
Fingerprint
Dive into the research topics of 'Certified safe real-time control of cooperative aerial robot systems with explicit reference governors'. Together they form a unique fingerprint.-
VLAAI1: Flanders Artificial Intelligence Research program (FAIR) – second cycle
Nowe, A. (Administrative Promotor) & Vanderborght, B. (Co-Promotor)
1/01/24 → 31/12/28
Project: Applied
-
FWOSB33: Prosthesis for Energy Efficient Walking and Running Powered by Redundant Series-Parallel Elastic Actuation
Vanderborght, B. (Administrative Promotor) & Convens, B. (Mandate)
1/01/17 → 31/12/20
Project: Fundamental
Cite this
- APA
- Author
- BIBTEX
- Harvard
- Standard
- RIS
- Vancouver