https://www.mdu.se/

Recent industrial developments in autonomous systems, or agents, which assume that humans and the agents share the same space or even work in close proximity, open for new challenges in robotics, especially in motion planning and control. In these settings, the control system should be able to provide these agents a reliable path following control when they are working in a group or in collaboration with one or several humans in complex and dynamic environments. In such scenarios, these agents are not only moving to reach their goals, i.e., locations, they are also aware of the movements of other entities to find a collision-free path. Thus, this paper proposes a dependable, i.e, safe, reliable and effective, path planning algorithm for a group of agents that share their working space with humans. Firstly, the method employs the Theta* algorithm to initialize the paths from a starting point to a goal for a set of agents. As Theta* algorithm is computationally heavy, it only reruns when there is a significant change of the environment. To deal with the movements of the agents, a static flow field along the configured path is defined. This field is used by the agents to navigate and reach their goals even if the planned trajectories are changed. Secondly, a dipole field is calculated to avoid the collision of agents with other agents and human subjects. In this approach, each agent is assumed to be a source of a magnetic dipole field in which the magnetic moment is aligned with the moving direction of the agent. The magnetic dipole-dipole interactions between these agents generate repulsive forces to help them to avoid collision. The effectiveness of the proposed approach has been evaluated with extensive simulations. The results show that the static flow field is able to drive agents to the goals with a small number of requirements to update the path of agents. Meanwhile, the dipole flow field plays an important role to prevent collisions. The combination of these two fields results in a safe path planning algorithm, with a deterministic outcome, to navigate agents to their desired goals.

In this paper, a novel path planning algorithm for multiple robots using congestion analysis and control is presented. The algorithm ensures a safe path planning solution by avoiding collisions among robots as well as among robots and humans. For each robot, alternative paths to the goal are realised. By analysing the travelling time of robots on different paths using Petri Nets, the optimal configuration of paths is selected. The prime objective is to avoid congestion when routing many robots into a narrow area. The movements of robots are controlled at every intersection by organising a one-by-one passing of the robots. Controls are available for the robots which are able to communicate and share information with each other. To avoid collision with humans and other moving objects (i.e. robots), a dipole field integrated with a dynamic window approach is developed. By considering the velocity and direction of the dynamic obstacles as sources of a virtual magnetic dipole moment, the dipole-dipole interaction between different moving objects will generate repulsive forces proportional to the velocity to prevent collisions. The whole system is presented on the widely used platform Robot Operating System (ROS) so that its implementation is extendable to real robots. Analysis and experiments are demonstrated with extensive simulations to evaluate the effectiveness of the proposed approach.

Path finding for multiple robots is one of most important problems in robotics when to find a way to move robots from their starting positions to reach their respective goals without collisions. However, in the case of a complex environment with the presence of humans and other unpredictable moving objects, fixing a single path to the goal may lead to a situation where there are a lot of obstacles on the planned path and the robots may fail to realise the moving plan. To address this issue, a new approach of using multiple path planning where each robot has different options to choose its path to the goal is introduced in this paper. The information about planned moving paths are shared among the robots in the working domain, combined with obstacle avoidance constraints in local ranges, and formulated as an optimisation problem. Solution of the problem leads to the optimal moving plans of robots. The effectiveness of the proposed approach is demonstrated by experimental results.