2000 Fiscal Year Final Research Report Summary
Experimental Study on Real World Mobile Robots which Operate in Realtime Considering Uncertainties in Perception and Action and Computational Cost
| Project/Area Number |
10558050
|
|---|
| Research Category |
Grant-in-Aid for Scientific Research (B).
|
| Allocation Type | Single-year Grants |
|---|
| Section | 展開研究 |
|---|
| Research Field |
Intelligent informatics
|
|---|
| Research Institution | Osaka University |
|---|
Principal Investigator |
MIURA Jun Graduate School of Engineering, Osaka University Associate Professor, 大学院・工学研究科, 助教授 (90219585)
|
|---|
| Co-Investigator(Kenkyū-buntansha) |
SHIMADA Nobutaka Graduate School of Engineering, Osaka University Research Associate, 大学院・工学研究科, 助手 (10294034)
KUNO Yoshinori Department of Engineering, Saitama University Professor, 工学部, 教授 (10252595)
SHIRAI Yoshiaki Graduate School of Engineering, Osaka University Professor, 大学院・工学研究科, 教授 (50206273)
勞 世〓 オムロン株式会社, 技術本部・係長(研究職)
TSUTUMI Yasuhiro Business Development Group, OMRON Corporation Senior Engineer, 事業開発本部, 係長(研究職)
LAO Shihong Information Technology Research Center, OMRON Corporation Senior Engineer
|
|---|
| Project Period (FY) |
1998 – 2000
|
| Keywords | Moible robot / planning under uncertainty / uncertainties in vision and motion / environment modeling under uncertainty / landmark extraction and selection / planning in dynamic environments / parallel scheduling / realtime omnidirectional stereo |
|---|
| Research Abstract |
The results of this research are summarized as follows :
1. Planning considering planning cost : Planning under uncertainty is generally costly because many possible outcomes of actions should be considered. We have developed an planning algorithm which, by explicitly considering the tradeoff between the planning cost and the plan quality, adaptively allocates the time for planning to minimize the total cost of planning generation and plan execution.
2. Parallel scheduling of multiple activities : Robots are usually composed of multiple processing units (called activities). To increase the efficiency of such a parallel system, we have developed a parallel scheduling method which is based on the dependency relationship between activities.
3. Environment modeling and navigation : Sensor-based environment modeling is necessary for safe movement of robots. We first examined the relationship between a robot task (what goal should be achieved in what kind of environment) and an appropriate mode
… More
ling method. We then developed several modeling methods and navigation methods which considers various kinds of uncertainties in perception and action.
4. Environment modeling and planning in dynamic environments : In a dynamic environment, changes of the environment itself and the value of each action should be considered in planning. We have investigated the modeling of such changes. In addition, the severer time pressure requires reduction of computational costs. We have developed a three-level planning architecture which has the knowledge-based meth-planner at the highest level to reduce the planning cost in a heuristic way.
5. Realtime omnidirectional stereo for detection and tracking of moving obstacles : mobile robots need fast, wide-field-of-view, and reliable recognition of surrounding environments for safe and efficient navigation. We have developed an omnidirectional realtime stereo system which uses a pair of vertically aligned omnidirectional cameras. We have realized the detection and tracking of moving obstacles by a moving robot. Less
|
