2015 Fiscal Year Annual Research Report
実環境における歩行安定性と移動効率を考慮した２足歩行ロボットの歩行軌道計画
| Project/Area Number |
15J06497
|
|---|
| Research Institution | Waseda University |
|---|
Principal Investigator |
ブランダン マルティン 早稲田大学, 理工学術院, 特別研究員(DC2)
|
|---|
| Project Period (FY) |
2015-04-24 – 2017-03-31
|
| Keywords | motion planning / footstep planning / humanoid robots / friction estimaton / stereo vision |
|---|
| Outline of Annual Research Achievements |
1) I developed a footstep planning algorithm which considers energy consumption, terrain-friction and stability of a biped robot. The results were published in the proceedings of the IEEE Humanoids 2015 conference and presented on a workshop at the IEEE IROS 2015 conference. I obtained interesting results, such as both low electrical energy consumption and human-like paths on slopes and on environments with slippery patches, by a choice of objective functions and constraints inspired by human gait literature. I reduced expected electrical energy consumption up to 50% when compared to other objective functions common in the footstep planning literature.
2) I developed computer vision algorithms for estimation of friction from images, and I studied the error associated with the task: both in humans and with the computer vision algorithms. I obtained very interesting and important results, such as: humans are poor at estimating friction from vision, humans use gloss as a cue for friction estimation, material (e.g. wood, concrete) is the most effective feature for predicting friction.
3) I analyzed the performance of 3D reconstruction using stereo vision depending on the functions/parameters used for the stereo matching. I improved understanding of stereo matching and probabilistic occupancy grids. Results were published on the IEEE Transactions PAMI journal.
|
| Current Status of Research Progress |
Current Status of Research Progress
2: Research has progressed on the whole more than it was originally planned.
Reason
- Item 2) of "研究実績の概要" was initially part of the plan for 2016, but I moved it to 2015 for smoother research progress. In this sense the research is progressing faster than planned.
- Item 1) of "研究実績の概要" was initially planned to be evaluated on real robot experiments during 2015. Instead, experiments were conducted mainly in simulation, and I moved the bulk of real robot experiments to 2016. I made this decision consciously in order to have working perception algorithms (Item 2) before conducting these experiments.
- To summarize, overall the amount of progress is as expected, although part of the order of planned research was changed (item 1 and 2).
|
| Strategy for Future Research Activity |
- I will further extend the current locomotion planner in order to optimize whole-body motion (not just footsteps) using the same objective functions as the footstep planner.
- I will setup a challenging evaluation scenario in the laboratory with slopes and several surfaces of different friction.
- I will use the friction estimation algorithms and stereo vision algorithms already developed together with the planner on the real robot and challenging scenario setup. I will measure energy consumption of the real robot and evaluate the whole system in this setup.
|
