Abstract | ||
---|---|---|
Long-range navigation in unknown environments is one of the major challenges for autonomous robot. Traditional methods are usually relying on metric mapping and path planning, which own a very high calculation cost for planning, re-planning, map building and map updating. Q-learning can learn a reactive navigation behavior for mobile robot, but easy fall into local minimum with long-range goal. In order to accomplish the long-range navigation task, a two-layers navigation framework is based a novel topological mapping (topo-mapping) method and Q-learning. The proposed topo-mapping method can incrementally build a topological map (topo-map) in unknown environment. Topo-map maintains two types of nodes: station node and feature node. Station nodes store low resolution information of environment and its related feature nodes store the export information. Local target can be selected from those feature nodes by proposed topology planner. This local target guides Q-learning based robot reach long-range goal. We evaluate the proposed method in simulation environment. Experimental results show that the proposed method based autonomous robot can reach the long-range goal successfully in unknown environment. |
Year | DOI | Venue |
---|---|---|
2019 | 10.1109/RCAR47638.2019.9044156 | 2019 IEEE International Conference on Real-time Computing and Robotics (RCAR) |
Keywords | DocType | ISBN |
autonomous robot,metric mapping,path planning,map building,reactive navigation behavior,long-range navigation task,two-layers navigation framework,topo-mapping method,topological map,topo-map,unknown environment,station node,feature node,Q-learning based robot | Conference | 978-1-7281-3727-8 |
Citations | PageRank | References |
0 | 0.34 | 8 |
Authors | ||
4 |
Name | Order | Citations | PageRank |
---|---|---|---|
Yanming Hu | 1 | 0 | 0.34 |
Decai Li | 2 | 0 | 0.34 |
Yuqing He | 3 | 63 | 19.58 |
Jianda Han | 4 | 220 | 60.61 |