The project aims through experimental research develop technologies to build private 5G networks for robotics missions under disadvantage conditions complementary functional real-time coupling existing (public) 5G and also 4G/LTE, WiFI, LoRaWAN networks. We propose to develop optimization and control algorithms of the communications and efficient data sharing in mission control using a team of mobile robots performing regular tasks such as pick-and-delivery and perimeter patrolling. We plan to analyze the connectivity behavior of mobile robots in the field and propose solutions to avoid connection loss and reestablish a connection autonomously. Further, we propose to address critical emergencies through automated communication infrastructure building enhanced by environment modeling performed by mobile robots.

The Sub-project focuses on R&D in robotics under CAMAT, targeting two key areas - (i) robot design inspired by Physical Intelligence (PI), and (ii) advanced control using Machine Learning (ML) and Artificial Intelligence (AI). It aims to transfer fundamental research into innovative technologies with high interdisciplinary impact. Key developments include a magnetic soft robotic gripper and micromanipulator for precise macro/micro manipulation, an articulated magnetic climbing robot with aerial deployment for inspection in hard-to-reach areas, and a shape-adaptive soft robot for confined spaces. The project also explores ML-based methods for fast deployment and control of complex robotic systems, and develops strategies for multicontact locomotion by estimating external wrenches critical to human-robot interaction and terrain traversal.

The aim of the ROBOPROX project is to establish a foundation for the growth of cutting-edge research in the fields of robotics and industrial manufacturing by supporting excellent research teams in this domain, expanding current knowledge, creating original inventions, strengthening the academic programs of project partners, and leveraging both national and international collaboration in top-tier research. The goal is to transform Czech and European industries to deliver more flexible, complex, competitive, and sustainable industrial production. Mathematical modeling, data-driven approaches, simulation, optimization algorithms, and formal methods are increasingly popular in industry. However, there is a lack of fundamental concepts, interoperable models, and high-performance algorithms—highlighting the need for suitable tools. This need presents a clear opportunity for ROBOPROX to create an outstanding new research environment for developing and deploying innovative research approaches in the industrial sector.

The project focuses on innovative methods for improvement of critical infrastructure (CI) security using digital models, IoT sensors, automated robots and data fusion and decision support systems. The structure of the system will be based on the concept of 5C architecture for cyber-physical systems to combine and interconnect heterogeneous systems, consisting of unattended sensors (personal ID cards, environmenal sensors, cameras), mobile robots (drones, wheeled and walking robots, ships), communication infrastructure, and algorithms for data collection, fusion and evaluation. Operation automation and the use of an event management system will improve situational awareness and operator’s decision making support. The properties of the system will be verified in a real environment.

In the project, we aim to address motion planning for robotic systems such as multi-legged walking in complex tasks of traversing unstructured terrain or climbing a wall. The proposed approach is to exploit the capabilities of learnable locomotion controllers to build a set of locomotion skills applicable in complex motion tasks. We plan to investigate biologically inspired locomotion control based on the coupling of neural oscillators to develop neural-based locomotion controllers with high plasticity capable of learning multiple gaits. We further plan to employ precise motion planning to synthesis motion planning policies using learning techniques such as hierarchical temporal memory. Complementary, we propose to combine global, less accurate models with a more precise local model using inter-basin actions and employ methods of deep reinforcement learning to provide smooth transitions between locally stable regions of individual motion controllers. We aim to establish complex analyses of the proposed approaches and experimentally verify them in scenarios with real robotic systems.

In the project, we aim to establish theoretical foundations for solving robotic routing problems with continuous optimization. We plan to leverage existing theoretical results on lower bound estimations of the close enough traveling salesman problem and model-based multi-goal motion planning for the Dubins vehicle model towards a general solution to combinatorial routing with continuous optimization problems arising in robotic scenarios. We target to establish algorithmic foundations for solution quality estimations based on tight lower bounds that can be employed in efficient pruning of search space and thus support finding optimal solutions. We target to 1) optimal solutions of routing problems with cost corresponding to the continuous optimization of multivariable functions arising from limitations of robotic systems; 2) design new efficient algorithms with machine learning-enabled scalability to solve large practical instances of realistic problems; 3) establish complex analysis and empirical evaluation of the developed solutions in experimental scenarios with robotic systems.

The goal of the project is to develop a multi-goal planning framework to provide optimal or near-optimal curvature-constrained tours in robotic applications particularly motivated by mission planning for unmanned aerial vehicles and motion planning of steerable needle. Through the project objectives, we aim to deliver a computationally efficient solution to the particular robotic variants of the traveling salesman problem in which individual tours respect motion constraints of real robotic systems. In particular, we will focus on relatively small instances of the computationally challenging planning problem with the specific aim to deliver tight lower bounds and solution quality estimation.

Further commercial deployment of small UAS requires addressing the operation safety and thus ensure safe flight execution together with the minimization of the risk to other air traffic and people and properties on the ground. Furthermore, the deployment needs automation of operation to decrease the operators´ workload and thus allow the operator to control multiple aircraft simultaneously and decrease operating costs. The project aims to (i) develop a general policy for risk-aware trajectory planning considering the off-nominal situations, (ii) to reduce the risk induced by off-nominal situations by automating decision making in such situations, (iii) automate landing site detection by proper mission planning, and (iv) to develop a controller capable of reliable autonomous landing.