In context
In 2023, autonomous exploration remained a bottleneck for mobile robots in industrial and service applications. Traditional SLAM required manual control for map building, which was inefficient and impractical in inaccessible areas. Researchers at Fudan University addressed this by enhancing frontier-based exploration with topological memory to reduce redundant revisits.
What was reported
The team proposed TMRRT (topological map based rapidly exploring random tree), which integrates variable-growth-rate local and global RRTs for frontier detection, Mean Shift clustering to reduce frontier data, and a topological map that stores historical best exploration points. This map guides the robot to avoid previously explored regions, reducing loop-closure inefficiencies.
In simulations and real-world tests, TMRRT reduced average exploration time by over 7.5% and path length by over 19.8% compared to standard RRT, and by over 15.7% and 34.3% respectively compared to frontier-based approach (FA). The algorithm dynamically updates the topological map using BFS to compute distances, ensuring efficient traversal.
The method is designed for indoor environments and is compatible with common SLAM frameworks, offering a practical solution for autonomous map building without manual intervention.
Why it mattered
This work demonstrated that adding a topological memory layer to sampling-based exploration can significantly improve efficiency, a key step toward fully autonomous mobile robots in dynamic industrial settings. Reduced path length and time translate to lower operational costs and faster deployment in warehouses, factories, and other facilities.
“The results show that the algorithm can effectively improve the efficiency of robot autonomous exploration and is feasible in the actual environment.”
Source: 《机器人》期刊 (robot.sia.cn) · Published 2023-05-09 · “基于拓扑地图的移动机器人室内环境高效自主探索算法”