Research On Key Techniques For Boundary Detection With Mobile Robots

With the rapid development of embedded computing and wireless communica-tion technologies, robots are more and more widely used to perform a great variety of tasks. As one significant part of such applications, robot boundary detection can provide useful boundary information for various application fields such as scientific research, military and business applications. Such boundary information gives those applications the ability to navigate, self-localize and fulfill specific searching task more effectively.However, boundary detection with mobile robots is a critical challenge due to the complexity and variety of the physical environment and the limitation of robot devices.First, to effectively and accurately sense the boundary information is difficult, not only due to the existing errors in sensors which are equipped on robots, but also due to the impact of noisy environmental data.Boundary conditions are becoming dynamic and open which further increase the complexity of tracking boundary. Especially due to the dynamism of the physical environment,applications needs to be aware of the static properties of the environment. Moreover, the temporal evolution of environment status also needs to be captured. All the issues above make it challenging to detect boundary with mobile robots.Second, one single robot can only cover limited area due to its limited resources. Single robot systems also suffer from the single point of failure. Thus, multi-robot systems are typically used for boundary detection. When multiple robots are used for boundary detection, collision between robots may frequently occur, but this issue has not been sufficiently addressed in the existing work.To address these challenges, this work mainly focuses on boundary detection with multiple mobile robots. Specifically, the main contributions of this work include:First, predicate detection is applied to effectively detect environmental boundary properties. In order to deal with the dynamic boundary properties, we begin with mod-eling of the static boundary data. Then snapshot predicates are specified to delineate current boundary states. Static boundary properties can be monitored by detecting snapshot predicates. We further model the temporal evolution of boundary data, and specify sequence predicates to delineate temporal properties of boundary data evolu-tion. Dynamic properties of the boundary can be monitored by detecting sequence predicates. We carry out the predicate detection method on a practical robot car to detect the boundary of a rectangular region.Second, we propose the method of anti-collision detection and leader-election based collision avoidance to solve the collision problem during cooperative boundary detection. Through anti-collision detection, we can predict in advance the possible collision. Leader election is conducted to determine a robot as the coordinator, so that the other robots can follow the order from this elected leader to avoid possible collision. Finally, we apply our proposed method to a practical boundary detection experiment with two actual robot cars. The experiment results show the effectiveness of our method successfully in coping with collision problem.