This paper presents a novel framework based on edge computing, implemented using Kubernetes orchestration, to optimally offload the computational tasks required for centralized control of multiple robotic agents. Edge-based centralized control architectures are prone to failure due to communication delays. The proposed framework computes the maximum round-trip time delay for which the system remains stable and modifies the controller parameters to ensure the control computation within the critical time. For higher processing and communication delays, the complexity of the controller needs to be reduced by reducing the number of agents, the prediction horizon, and the efficient use of edge resources. The edge resources are dynamic, and the controller needs to be designed to guarantee the online computation within a desired time. A dynamic resource allocation method (based on an approximate function of the controller parameters, complexity, and computational resources) is proposed to design the controller parameters to ensure the bounded computation time. To validate the effectiveness of the proposed approach, we conduct experimental evaluations that analyze system behavior under various conditions, providing valuable insights into the performance, scalability, and robustness of multi-agent control systems deployed on edge infrastructure.

In this article, we present the Layered Semantic Graphs (LSG), a novel actionable hierarchical scene graph, fully integrated with a multi-modal mission planner, the FLIE: A First-Look based Inspection and Exploration planner [1]. The novelty of this work stems from aiming to address the task of maintaining an intuitive and multi-resolution scene representation, while simultaneously offering a tractable foundation for planning and scene understanding during an ongoing inspection mission of apriori unknown targets-of-interest in an unknown environment. The proposed LSG scheme is composed of locally nested hierarchical graphs, at multiple layers of abstraction, with the abstract concepts grounded on the functionality of the integrated FLIE planner. Furthermore, LSG encapsulates real-time semantic segmentation models that offer extraction and localization of desired semantic elements within the hierarchical representation. This extends the capability of the inspection planner, which can then leverage LSG to make an informed decision to inspect a particular semantic of interest. We also emphasize the hierarchical and semantic path-planning capabilities of LSG, which could extend inspection missions by improving situational awareness for human operators in an unknown environment. The validity of the proposed scheme is proven through extensive evaluations of the proposed architecture in simulations, as well as experimental field deployments on a Boston Dynamics Spot quadruped robot in urban outdoor environment settings.

Cloud robotics has emerged as a promising technology for robotics applications due to its advantages of offloading computationally intensive tasks, facilitating data sharing, and enhancing robot coordination. However, integrating cloud computing with robotics remains a complex challenge due to network latency, security concerns, and the need for efficient resource management. In this work, we present a scalable and intuitive framework for testing cloud and edge robotic systems. The framework consists of two main components enabled by containerized technology: (a) a containerized cloud cluster and (b) the containerized robot simulation environment. The system incorporates two endpoints of a User Datagram Protocol (UDP) tunnel, enabling bidirectional communication between the cloud cluster container and the robot simulation environment, while simulating realistic network conditions. To achieve this, we consider the use case of cloud-assisted remote control for aerial robots, while utilizing Linux-based traffic control to introduce artificial delay and jitter, replicating variable network conditions encountered in practical cloud-robot deployments,

This article presents a novel edge-based architecture to dynamically allocate resources to edge-offloaded controllers for multi-agent systems. The proposed controllers are designed to generate collision-free trajectories to track the desired reference positions. The computational complexity of the controllers’ problem is estimated by a second-order polynomial regression model, while the Least Squares (LS) minimization technique is employed for the coefficients’ estimation. The covariance matrix plays an essential role in assessing the confidence in the parameter estimates and in investigating correlations among parameters. Through this curve fitting process, we can dynamically estimate the complexity of the controllers’ problem as conditions change, enabling effective and responsive resource allocation. Furthermore, a novel control law is designed to control the dynamic resource allocation, based on the measured communication and processing time delays. This approach allows us to control the controllers’ response time, thereby ensuring the closed-loop system’s stability. The overall architecture is enabled through a Kubernetes (k8s) cluster and is experimentally evaluated.

Advancements in satellite technology have enabled the deployment of large constellations in low Earth orbit (LEO), presenting significant challenges in orbital management. Operating hundreds or even thousands of satellites imposes considerable computational and communication demands on ground control systems and operators. A primary operational challenge is maintaining satellites on their nominal trajectories in the presence of continuous perturbations, such as atmospheric drag and third-body gravitational forces. This article presents a reinforcement learning (RL)-based approach for decentralized satellite station keeping (SK). The proposed method utilizes a neural network policy—trained with the model-free soft actor-critic (SAC) algorithm in a high-fidelity, physics-based simulation environment—to compute corrective thrust vectors based on observed deviations from the satellite’s nominal unperturbed trajectory. The resulting policy is computationally efficient and suitable for deployment onboard resource-constrained, space-grade systems. The performance of the RL-based controller is evaluated through Monte Carlo simulations and compared with that of a conventional linear model predictive controller (MPC), which is widely adopted due to its relatively low computational requirements. The comparison focuses on computational complexity, control performance, and robustness. Results demonstrate that the RL-based controller can achieve improved maneuver efficiency by directly learning the nonlinear relative dynamics, without incurring the computational cost typically associated with classical nonlinear optimization-based control methods. These findings underscore the potential of RL techniques for scalable and autonomous management of satellite constellations.

Advancements in satellite technology have led to the development of large constellations in Low Earth Orbit, which presents new challenges in orbital management. Controlling and managing these large numbers of satellites efficiently becomes un-scalable due to the high computational and telemetry demands. This article addresses the problem of station-keeping, where each satellite in a constellation independently calculates correction maneuvers to compensate deviations from the nominal orbit caused by orbital perturbations. Using Reinforcement Learning, a decentralized satellite station-keeping policy is trained in a high-fidelity simulation environment, to output a low-thrust finite-pulse maneuver plan. The trained neural network policy requires minimal computational resources and can be readily deployed onboard resource-constrained space-grade computers. The proposed framework employs the model-free Soft Actor-Critic algorithm, which observes the relative orbital elements between the satellite’s current and ideal/desired trajectory and outputs a maneuver plan consisting of thrust direction, start time, and duration. Two policies are trained to account for both in-plane and out-of-plane tracking. To this end, a realistic and fuel-efficient mission scenario is designed, keeping orbit-plane errors within specified bounds. Furthermore, the performance of the proposed framework is compared with an optimal-control-based station-keeping approach. The efficacy and robustness of the proposed framework is demonstrated through a series of Monte-Carlo simulations and benchmarked against the traditional optimization-based approach, on a wide array of initial conditions.

In this article, we design an adaptive controller for the position and heading control for a payload-carrying spacecraft to perform docking with a target docking station. We address the problem by identifying the state constraints required to safely dock the spacecraft and imposing these constraints on an adaptive tracking controller. To make the controller adapt to different types of payloads, the adaptive controller is designed without any explicit a priori knowledge of the system dynamics or bound for the uncertainties. Furthermore, to accommodate a wide range of initial conditions, the constraints are chosen to be time-varying. Thus, unlike conventional controllers, the proposed controller enforces the safety of the spacecraft during docking by imposing state constraints while adapting to unknown drastic dynamic variations. The controller is validated in simulation for docking a 6 DoF spacecraft in the orbital space. Additionally, for technology readiness, we have performed the hardware validation of the controller using a payload-carrying planar floating robot and a prototype docking station. Compared to the state-of-the-art controllers, the proposed controller guarantees predefined time-varying state constraints while significantly improving the performance.