Multi-agent 3D exploration in dynamic environments

Sammanfattning

Efficient exploration and mapping of dynamic 3D environments are critical for multiagent autonomous systems, especially in robotics and related domains. There are manydifficulties when working within such a domain, some of which are planning safe routes,avoiding collisions with objects and agents, efficiently mapping the environment and coverage of the entire area. In order to solve these issues there has to be a planner that can handle dynamic obstacles, 3D environments and multi-agent scenarios. As of now, there seemsto be a lack of research done within this combined problem domain, but much progresshas been made within the isolated domains (i.e. dynamic obstacles, 3D environments,multi-agent). This thesis aims to merge the recent developments made within these domains in order to solve the combined problem statement, by adapting and building uponthe solutions within each field. The proposed planner, MADAEP, Multi-Agent DynamicAutonomous Exploration Planner has been evaluated against current state-of-the-art exploration planners for multi-agent systems. The planner builds upon the single-agent planner Dynamic Autonomous Exploration Planner (DAEP) with improvements and multiagent functionality. Results indicate that MADAEP achieves a similar or higher amount oftotal coverage after exploration has been completed, and also scales well with an increasein dynamic obstacles and larger exploration areas.