An improved A* algorithm for searching the minimum dose path in nuclear facilities

Abstract

Nuclear power plant workers are inevitably exposed to radiation during routine maintenance work. Therefore, optimization of the walking-path is necessary to reduce the exposure radiation. However, the complex environment of nuclear power plants makes it hard to find the path with the minimum cumulative radiation dose. An improved A* algorithm is proposed in this paper, and it is more advantageous than the traditional A* algorithm. It can not only be suitable for more complicated environments but also obtain a more optimized path. Through the Unity 3D platform, which is an efficient game engine, the simulation experiments are performed based on the environment of a virtual factory. The simulation results show that the improved A* algorithm for searching the minimum dose path is more reliable and more effective. It provides a route with a lower cumulative dose, and the path length is appropriate. In the complex environment, the cumulative dose calculated by the improved algorithm is much smaller than that calculated by the traditional algorithm. Improved A* algorithm has a strong anti-interference ability and performs well in complex environments. It can provide path-planning for nuclear facility staff to reduce radiation exposure.

Introduction

The nuclear radiation protection of nuclear power plants is the priority for the normal operation. Even though nuclear power plants have always had many nuclear radiation protection measures, the workers would still be exposed to the radioactive environment. In particular, when an accident occurs due to human factors, technical factors, or natural disasters, the staff would face serious threats. Hence, it is a necessity to provide an appropriate route for staff to reduce the radiation they receive, which is consistent with the ALARA (as low as reasonably achievable) principle.

The importance of path-planning is becoming increasingly prominent. Alzalloum (2009) solved the lowest cost path problem in radiologically contaminated areas by using the Bellman-Ford and Dijkstra algorithms to find the minimum dose path. Khasawneh et al. (2013a, b) expected to obtain the useful information for path-planning through an excellent distributed wireless sensing infrastructure and proposed a local navigation algorithm that was tested by the “Radiation Evasion” and the “Nearest Exit” criteria. Chao et al. (2018) introduced a sampling-based method that fully utilized effective information and their method performed well in both dynamic and static environments. In order to deal with the problems of path-planning, Wang and Cai (2018b) also established a useful model for dose calculation based on particle swarm optimization algorithm. Liu et al. (2014) conducted research on path-finding based on particle swarm optimization (PSO) algorithm and solved the minimum dose navigation problem of the multi-objective path. Further, when facing the problem of multiple radiation regions, Liu et al. (2016) designed an effective method for path-planning. Considering that the radiation field is dynamic, Li et al. (2016) also proposed an appropriate method to cope with path-planning problems in the dynamic environment.

The A* algorithm is a typical graph search algorithm, and it has widely been used. Compared with the traditional A* algorithm, a probabilistic roadmap method was proposed by Wang and Cai (2018a), and it is proved to be more suitable for emergencies. Besides, Liu et al. (2015) used the A* algorithm based on a road network and established a minimum dose method. Cui and Shi (2011) revealed the relationships between the A*-based algorithms in modern computer games from various perspectives. In addition, Persson and Sharf (2014) used the classic A* algorithm to present a class of sampling-based path-planning algorithms.

It's also worth mentioning that virtual simulation is an essential part of algorithm research. Excellent virtual simulation can not only effectively complete the path-planning work, but also make it directly perceived through the senses. Chao et al. (2017) proposed a sampling-based method (SBM). Compared with the Dijkstra algorithm, SBM showed its superiority visually through virtual reality technology. The game engine is a good choice for virtual simulation. Jacobson and Lewis (2005) introduced the virtual reality based on CaveUT, which was a freeware project. Mol et al. (2009) used a game engine to create a virtual environment and got satisfactory results.

Among the current virtual simulation platforms, Unity 3D is powerful with many appropriate APIs (Application Programming Interfaces). For example, Raycast casts a ray that can get information back from the specific objects, allowing us to interact with the environment. Besides, The use of Collider makes collision detection easy to implement. We thus use Unity 3D to carry out the virtual simulations. The key to the A* algorithm is the setting of the heuristic function. The heuristic function significantly affects the efficiency and accuracy of the A* algorithm, and It restricts the applicable scope of the algorithm. In the nuclear facility, the environment is complex, which makes the heuristic function of the traditional A* algorithm less effective. According to the result of the experimental simulation, The heuristic function is easily affected by the complex environment when the cost between nodes is evaluated by a straight-line path. In this paper, we try to propose an improved A* algorithm by improving the heuristic function. The improved heuristic function has higher efficiency and accuracy by appropriately enlarging the search area. What is more, it is flexible enough for a broader range of environments. In this paper, virtual simulations are performed based on the Unity 3D platform. Our results show that the improved A* algorithm is more practical by improving its heuristic function.

The structure of this paper is given as follows. Section 2 briefly introduces the A* algorithm. Section 3 introduces the construction of simulation environments and focuses on the discussion of the difference between these two A*-based algorithms. Section 4 conducts the simulation experiments and discusses the results. At last, Section 5 makes a conclusion.