Browsing by Author "Niu, Hanlin"

Now showing 1 - 5 of 5
  • Thumbnail Image
    ItemOpen Access
    Consensus-based deep reinforcement learning for mobile robot mapless navigation
    (IEEE, 2024-06-05) Liu, Wenxing; Niu, Hanlin; Caliskanelli, Ipek; Xu, Zhengjia; Skilton, Robert
    When using mobile robots to perform data collection about the surroundings, the performance might be dissatisfying since the environments could be unknown and challenging. This situation will pose challenges for mobile robot navigation and exploration. To tackle this issue, we propose a consensus-based deep reinforcement learning (DRL) algorithm for multiple robots to perform mapless navigation and exploration. The proposed algorithm leverages both consensus-based training and DRL, which reduces required training steps while maintaining the same training reward. Once trained with fixed obstacles, the proposed training model can demonstrate adaptability in handling real-world random static obstacles and sudden obstacles. The experimental video is available at: at: https://youtu.be/ym2yvbKg4fU.
  • Thumbnail Image
    ItemOpen Access
    Efficient path planning algorithms for Unmanned Surface Vehicle
    (Elsevier, 2016-11-01) Niu, Hanlin; Lu, Yu; Savvaris, Al; Tsourdos, Antonios
    The C-Enduro Unmanned Surface Vehicle (USV) is designed to operate at sea for extended periods of time (up to 3 months). To increase the endurance capability of the USV, an energy efficient path planning algorithm is developed. The proposed path planning algorithm integrates the Voronoi diagram, Visibility algorithm, Dijkstra search algorithm and takes also into account the sea current data. Ten USV simulated mission scenarios at different time of day and start/end points were analysed. The proposed approach shows that the amount of energy saved can be up to 21%. Moreover, the proposed algorithm can be used to calculate a collision free and energy efficient path to keep the USV safe and improve the USV capability. The safety distance between the USV and the coastline can also be configured by the user.
  • Thumbnail Image
    ItemOpen Access
    Energy efficient path planning and model checking for long endurance unmanned surface vehicles.
    (2017-09) Niu, Hanlin; Savvaris, Al; Tsourdos, Antonios
    In this dissertation, path following, path planning, collision avoidance and model checking algorithms were developed and simulated for improving the level of autonomy for Unmanned Surface Vehicle (USV). Firstly, four path following algorithms, namely, Carrot Chasing, Nonlinear Guidance Law, Pure pursuit and LOS, and Vector Field algorithms, were compared in simulation and Carrot Chasing was tested in Unmanned Safety Marine Operations Over The Horizon (USMOOTH) project. Secondly, three path planning algorithms, including Voronoi-Visibility shortest path planning, Voronoi-Visibility energy efficient path planning and Genetic Algorithm based energy efficient path planning algorithms, are presented. Voronoi-Visibility shortest path planning algorithm was proposed by integrating Voronoi diagram, Dijkstra’s algorithm and Visibility graph. The path quality and computational efficiency were demonstrated through comparing with Voronoi algorithms. Moreover, the proposed algorithm ensured USV safety by keeping the USV at a configurable clearance distance from the coastlines. Voronoi-Visibility energy efficient path planning algorithm was proposed by taking sea current data into account. To address the problem of time-varying sea current, Genetic Algorithm was integrated with Voronoi-Visibility energy efficient path planning algorithm. The energy efficiency of Voronoi-Visibility and Genetic Algorithm based algorithms were demonstrated in simulated missions. Moreover, collision avoidance algorithm was proposed and validated in single and multiple intruders scenarios. Finally, the feasibility of using model checking for USV decision-making systems verification was demonstrated in three USV mission scenarios. In the final scenario, a multi-agent system, including two USVs, an Unmanned Aerial Vehicle (UAV), a Ground Control Station (GCS) and a wireless mesh network, were modelled using Kripke modelling algorithm. The modelled uncertainties include communication loss, collision risk, fault event and energy states. Three desirable properties, including safety, maximum endurance, and fault tolerance, were expressed using Computational Tree Logic (CTL), which were verified using Model Checker for Multi-Agent System (MCMAS). The verification results were used to retrospect and improve the design of the decision-making system.
  • Thumbnail Image
    ItemOpen Access
    An energy-efficient path planning algorithm for unmanned surface vehicles
    (Elsevier, 2018-05-25) Niu, Hanlin; Lu, Yu; Savvaris, Al; Tsourdos, Antonios
    The sea current state affects the energy consumption of Unmanned Surface Vehicles (USVs) significantly and the path planning approach plays an important role in determining how long the USV can travel. To improve the endurance of the USV, an energy efficient path planning approach for computing feasible paths for USVs that takes the energy consumption into account based on sea current data is proposed. The approach also ensures that the USV remains at a user-configurable safety distance away from all islands and coastlines. In the proposed approach, Voronoi diagram, Visibility graph, Dijkstra's search and energy consumption function are combined, which allows USVs to avoid obstacles while at the same time using minimum amount of energy. The Voronoi-Visibility (VV) energy-efficient path and the corresponding shortest path were simulated and compared for ten missions in Singapore Strait and five missions for islands off the coast of Croatia. Impact of parameters such as mission time, the USV speed and sea current state on the results were analysed. It is shown that the proposed VV algorithm improves the quality of the Voronoi energy efficient path while keeping the same level of computational efficiency as that of the Voronoi energy efficient path planning algorithm.
  • Thumbnail Image
    ItemOpen Access
    Verifying collision avoidance behaviours for unmanned surface vehicles using probabilistic model checking
    (Elsevier, 2016-09-30) Lu, Yu; Niu, Hanlin; Savvaris, Al; Tsourdos, Antonios
    Collision avoidance is an essential safety requirement for unmanned surface vehicles (USVs). Normally, its practical verification is non-trivial, due to the stochastic behaviours of both the USVs and the intruders. This paper presents the probabilistic timed automata (PTAs) based formalism for three collision avoidance behaviours of USVs in uncertain dynamic environments, which are associated with the crossing situation in COLREGs. Steering right, acceleration, and deceleration are considered potential evasive manoeuvres. The state-of-the-art prism model checker is applied to analyse the underlying models. This work provides a framework and practical application of the probabilistic model checking for decision making in collision avoidance for USVs.