2017 Fiscal Year Research-status Report
簡潔な・利用しやすい構造を有する学習ネットワークの構成と応用に関する研究
| Project/Area Number |
15K18095
|
|---|
| Research Institution | Waseda University |
|---|
Principal Investigator |
Parque Victor 早稲田大学, 理工学術院, 講師 (50745221)
|
|---|
| Project Period (FY) |
2015-04-01 – 2019-03-31
|
| Keywords | succinctness / graph representation / path planning / control / computer-aided design / evolutionary computing |
|---|
| Outline of Annual Research Achievements |
In 2017, the algorithmic foundations and application benchmarks for Genetic Counterfactual Programming (GCP) in computer-aided design and robotics were further developed: (1) computing succinct networks for a centralized graph in a polygonal map with convex and non-convex geometry, enabling the computation of efficient path planning and communication of multi-agent systems in cluttered environments; (2) the succinct representation of directed and undirected graphs with self-loops and its extensive evaluation which enlarges the scope and generalization ability of the conventional graph-representation schemes as well as realizes the succinct representation of graphs with self-loops via numbers; (3) the new methods for path bundling in bipartite networks using a compact encoding scheme, and its benchmarks with the state-of-the-art evolutionary computing algorithms through 7.5 billion shortest path computations, allowing the efficient coordination for the multi/compounded path planning of information, goods, and people; (4) the new method to avoid stagnation in evolutionary computing for control problems, which allows the efficient convergence towards the optimal solution, (5) the data-driven compounded path planning in which the need of computationally expensive controllers in trajectory following is completely avoided.
|
| Current Status of Research Progress |
Current Status of Research Progress
1: Research has progressed more than it was originally planned.
Reason
The algorithmic foundations and application benchmarks for Genetic Counterfactual Programming in Design and Robotics problems were further developed.
|
| Strategy for Future Research Activity |
In 2018, while deploying in real world systems, research on the applications of Genetic Counterfactual Programming are to be conducted: (1) Research on the structure of Genetic Counterfactual Programming, and (2) Research on the application of Genetic Counterfactual Programming in robot control and computer-aided design.
|
| Causes of Carryover |
The reason for Incurring Amount to be Used Next Fiscal Year is as follows: the acquisition of a computer suitable for sophisticated numerical computations was postponed to the consecutive year, since the existing hardware at Waseda University satisfies the computational loads to current date.
It is expected that the amount to be used in the next fiscal year is split among the travel costs and registration fees of conferences during 2018, the costs for acquiring a computer suitable for sophisticated numerical computations, relevant books, and publishing at relevant journals venues.
|
