Coverage Planning for UVC Irradiation: Robot Surface Disinfection Based on Swarm Intelligence Algorithm

Abstract

Ultraviolet (UV) radiation has been widely utilized as a disinfection strategy to effectively eliminate various pathogens. The disinfection task achieves complete coverage of object surfaces by planning the motion trajectory of autonomous mobile robots and the UVC irradiation strategy. This introduces an additional layer of complexity to path planning, as every point on the surface of the object must receive a certain dose of irradiation. Nevertheless, the considerable dosage required for virus inactivation often leads to substantial energy consumption and dose redundancy in disinfection tasks, presenting challenges for the implementation of robots in large-scale environments. Optimizing energy consumption of light sources has become a primary concern in disinfection planning, particularly in large-scale settings. Addressing the inefficiencies associated with dosage redundancy, this study proposes a dose coverage planning framework, utilizing MOPSO to solve the multi-objective optimization model for planning UVC dose coverage. Diverging from conventional path planning methodologies, our approach prioritizes the intrinsic characteristics of dose accumulation, integrating a UVC light efficiency factor to mitigate dose redundancy with the aim of reducing energy expenditure and enhancing the efficiency of robotic disinfection. Empirical trials conducted with autonomous disinfecting robots in real-world settings have corroborated the efficacy of this model in deactivating viruses.

Keywords: AIGaN-based UVC LED; coverage path planning; disinfection robot; irradiation planner.

Figure 1

Energy consumption of the robot chassis and UVC lamp set as a function of working time. (a) Illustration of irradiation of a robot equipped with UVC lamps. (b) As the working time varies, the energy consumption ratio between the UVC lamp group and the robot chassis is observed. It is evident that the UVC lamps account for a significant portion of the energy consumed for disinfection tasks.

Figure 2

(a) Schematic diagram illustrating the calculation method for robot-induced irradiation on object surfaces. (b) UVC irradiation dose map constructed based on 2D grid map.

Figure 3

(a) The side view of two adjacent waypoints, with waypoints set at different intervals and the robot statically irradiating at each waypoint. (b–d) LED static irradiation deployed at four waypoints, with varying distances between light sources, exploring the superimposed characteristics of irradiation at the waypoints. The UVC irradiation dose map is presented as a heatmap, and the doses obtained in each observation interval are also recorded.

Figure 4

Convergence comparison of five metaheuristic multi-objective optimization algorithms.

Figure 5

Metaheuristic algorithms addressing map tasks of varying areas to solve the Pareto optimal front of the irradiation multi-objective model. (a) Map area 15 m². (b) Map area 20 m². (c) Map area 25 m². (d) Map area 30 m².

Figure 6

The hierarchical framework of overall disinfection task planning, illustrating the execution flow of the irradiance planner and path planning algorithm. Given the initial population parameters for the robot, environment data, and three types of motion strategies, the best value is searched by simulating each individual within the population. MOPSO is executed to obtain waypoints and static dwell disinfection times. Sub-regions are divided based on the K-means algorithm, and finally, the TSP algorithm is used to determine the sequence of waypoints, generating an optimized trajectory.

Figure 7

Simulation of DCP planner in two structured obstacle terrains. (a,e) Waypoint locations after the dose coverage planner is executed by MOPSO (blue circle). (b,f) K-means clustering of waypoints, with division into cluster groups S₁ to S₄. (c,g) The clusters are ordered based on the robot’s starting location and cluster centroids, demonstrated in the highlighted yellow box. (d,h) Performing TSP on the sequence of waypoints within each group to obtain the ordered flight trajectories. Additionally, determining the starting point (green circle) and ending point (red square) of each subcluster based on the order of the groups.

Figure 8

Irradiation performance experiment. (a) AIGaN-based UVC LED chip, (b) UVC measurement probe, (c) constant current power supply, (d) irradiance measurement instrument, (e) light assembly with mounted AIGaN-based UVC LED chips.

Figure 9

Static irradiation characteristics curve of combined UVC light field. Experimental measurements of irradiation power from UVC light sets at different distances.

Figure 10

(a) The overall structure of the deep ultraviolet disinfection robot. (b) Distributed control structure and communication mechanism.

Figure 11

Autonomous disinfection experiment for mobile robots. (a) Laboratory scenario. (b) 2D grid map acquired after performing slam. (c,e) The marker location where the illuminometer is placed. (d) Floor plan of laboratory scene simulation, where orange dots represent illuminometer test points.

Figure 12

Disinfection in the laboratory and execution process of the planner: (a) The planner generates waypoints and corresponding timestamps. Triangles indicate the locations of the waypoints, and the shading represents the duration of disinfection at each waypoint. (b) K-means clustering divides the waypoints into regions. In addition, the areas marked in yellow indicate the execution order of movements between clusters, where c1 to c4 represent the centroids of each cluster. (c) The TSP algorithm is executed to generate the path.

Figure 13

Comparative experiment: The experiment compared the dose coverage effect after executing four path algorithms. (a,b) Spiral. (c,d) Voronoi. (e,f) APF. (g,h) Ours. The illumination energy measurement device was deployed at the observation points in Figure 8d, and the robot executed the trajectory according to the algorithm in the left figure. After the disinfection, the distribution of UVC dose at each point was measured as shown in the right figure. For easy comparison, the abscissa adopts a unified labeling order, and the ordinate adopts proportional scale.

Figure 14

Sensitive analysis of dose coverage planning algorithm to high- and low-dose targets. (a) Low-dose mode, 20 mJ/cm². (b) High-dose mode, 100 mJ/cm².