Unmanned aerial vehicle (UAV) swarms are increasingly used in critical applications such as aerial mapping, environmental monitoring, and autonomous delivery. However, the reliability of these systems is highly dependent on uninterrupted access to the Global Navigation Satellite Systems (GNSS) signals, which can be disrupted in real-world scenarios due to interference, environmental conditions, or adversarial attacks, causing disorientation, collision risks, and mission failure. This paper proposes SwarmRaft, a blockchain-inspired positioning and consensus framework for maintaining coordination and data integrity in UAV swarms operating under GNSS-denied conditions. SwarmRaft leverages the Raft consensus algorithm to enable distributed drones (nodes) to agree on state updates such as location and heading, even in the absence of GNSS signals for one or more nodes. In our prototype, each node uses GNSS and local sensing, and communicates over WiFi in a simulated swarm. Upon signal loss, consensus is used to reconstruct or verify the position of the failed node based on its last known state and trajectory. Our system demonstrates robustness in maintaining swarm coherence and fault tolerance through a lightweight, scalable communication model. This work provides a practical and secure approach for decentralized drone operation in unpredictable environments.

This paper presents experimental results of mobile robot navigation using two predictive controllers – a conventional model-predictive control and a Q-learning predictive controller. The latter essentially substitutes the running objective roll-outs with predicted action-value (Q-function) estimates. The idea behind such an approach is to integrate capabilities of reinforcement learning agents into the setting of model-predictive control while retaining the safety guarantees of the latter. Noteworthy the action sequence calculation step in both algorithms is of the same computational complexity. Yet, as we observed in our experiments, the learning predictive controller was able to outperform the model-predictive baseline. The code for the environment simulation may be found under https://github.com/thd-research/RL-autonomous-navigation.

Traditional control approaches for robotic systems, such as linear quadratic regulator (LQR) or model predictive control (MPC), often rely on a known model of the environment. However, in the real world, uncertainty is a common feature of control problems hence models have imperfections. In this work, we address reward engineering for underactuated robotic systems with parameter uncertainty. We introduce a novel reinforcement learning (RL) method to plan the reward function, specifically designed for underactuated robotic systems with parameter uncertainty. We present and validate a new algorithm called Greedy-Divide and Conquer. We implement this algorithm with a single RL agent to address the challenge of swinging up and balancing a Pendubot system with uncertain parameters and give another example with a 2D-Drone with body mass uncertainty. Our ultimate objective is to enhance the system’s ability to adapt and perform reliably in the face of varying uncertainties.

Slip control is a common and crucial functionality for a number of vehicles ranging from cars to tractors and trucks. The purpose of slip control is to improve vehicle's traction and motion stability, prevent excessive wheel slippage and provide stable braking. Slip is a non-linear function of the vehicle's ground speed and wheel rotation frequency and as such depends on the internal state variables, such as wheel load torque, which are in turn unknown. A number of approaches exist to estimate the unknown state, one of the most used ones being based on Kalman filter. In the current study, we present experimental results of slip control for an electrical single wheel-drive tractor using an unscented Kalman filter, which is a variant of Kalman filter suitable for non-linear systems. To cope with the problems of state estimation for heavy-duty vehicles, the Kalman filter was augmented with a fuzzy-logic supervisor aimed at assessment of vehicle dynamics. The goal of the supervisor was to adapt the state noise covariance with the goal of improving tracking accuracy. Wheel slip reduction was observed and its mean stayed within the desired limit. Experiments were carried out under two different road conditions and the condition change was identified by the Kalman filter

Electrified drive trains for tractors are supposed to realize great potential of raising performance in heavy operations via optimal traction control. The paper proposes to apply an adaptive unscented Kaiman filter (UKF) with a fuzzy supervisor for identification of electrical drive train tractor dynamics. The key advantage of electrical drive trains lies in feedback of drive torque which plays crucial role in traction parameter estimation. It is known that without using special adaptation techniques, an UKF may cause some divergence problems and lowered precision of estimation as well as its predecessor, an extended Kaiman filter (EKF). A method based on a fuzzy logic supervisor in addition to adaptation of an UKF is proposed to maintain trade-off between tracking strength and estimation accuracy. Simulation results with a comprehensive tractor dynamics model showed increase in estimation precision of traction parameters. Laboratory experiments using a test stand with an electrical load machine showed appropriate estimation of the load torque.