作業経路方向を考慮したレイアウト最適化のマニピュレータ軽量化のための複合領域設計最適化への影響の検証
Impact of Layout Optimization Considering Task Path Direction on Multidisciplinary Design Optimization for Manipulator Lightweighting

日比野圭歩, 遠藤央, 筒井幸雄 / Kaho Hibino, Mitsuru Endo, Yukio Tsutsui

日本機械学会 第15回最適化シンポジウム OPTIS2024


背景 – Background

本研究では,目的に最適なマニピュレータの設計を目的として,複合領域設計最適化(MDO:Multidisciplinary Design Optimization)を用いて軽量化を目指す.本研究ではこれまでに,6軸のマニピュレータを対象として,逐次最適化を用いて全体の総質量最小化を目的としてMDOに取り組んだ.このとき,可操作度最大となる設置位置と,軌道を実現するときの最も軽量となるモータと減速機を最適化するような2領域のMDOを構築した.また,設置位置最適化に用いていた可操作度[1]がスカラー量であることに着目し,経路方向を考慮した可操作性指標DMI(Directional Manipulability Index)を提案し,設置位置最適化での有効性を示した[2].本発表では,提案したDMIをMDOに導入し,全体の設計の最適化への有効性を検証し議論する.また,エネルギー指標として提案したDEI[3]もMDOに導入し,その有効性を確認する.

In this study, we aim to achieve lightweight design optimization of a manipulator using Multidisciplinary Design Optimization (MDO) to develop an optimal structure for a given objective. Previously, we applied MDO to a six-axis manipulator, employing a sequential optimization approach to minimize the system’s total mass.

To achieve this, we formulated a two-domain MDO framework: one for optimizing the installation position to maximize manipulability and another for selecting the lightest possible combination of motors and reducers that satisfy trajectory constraints. Additionally, we proposed the Directional Manipulability Index (DMI) as a new manipulability metric, incorporating directional effects along the trajectory to address the limitation of conventional scalar-based manipulability measures [1]. Our results demonstrated its effectiveness in installation position optimization [2].

In this presentation, we integrate the proposed DMI into the MDO framework and evaluate its effectiveness in the overall design optimization. Furthermore, we introduce the Directional Energy Index (DEI) [3] as an energy-based metric within MDO and validate its impact on optimization performance.


手法 – Method

1 複合領域設計最適化 / Multidisciplinary Design Optimization

複合領域設計最適化(MDO: Multidisciplinary Design Optimization)は,複合的なシステムの設計において,複数の設計領域で設計変数を同時に最適化することで包括的にシステム全体を最適化する手法である.本研究ではこれまでに,2領域を考慮したMDOに取り組んだ.図1に,本稿におけるMDOのアーキテクチャを示す.このMDOは,全体を最適化するシステムレベルオプティマイザ(SLO),幾何学的領域を考慮したサブオプティマイザ1(SO1),動力学的領域を考慮したサブオプティマイザ2(SO2)から構成される[4]. SLOでは軽量化を目的として前腕リンク・上腕リンクの長さ,SO1では可操作度の最大化を目的としてマニピュレータの設置位置,SO2では軽量化を目的として必要トルクと角速度を満たすモータ・減速機を最適化する. 本発表では,SO1の設置位置最適化にDMI,DEIを用いて,MDOでマニピュレータ全体を最適化する.

Multidisciplinary Design Optimization (MDO) is a method for comprehensively optimizing an entire system by simultaneously optimizing design variables across multiple design domains in complex system design. In this study, we have previously applied MDO considering two domains. Figure 1 illustrates the MDO architecture used in this work.

This MDO framework consists of a System-Level Optimizer (SLO) for overall optimization, Sub-Optimizer 1 (SO1)for geometric considerations, and Sub-Optimizer 2 (SO2) for dynamic considerations [4].

The SLO aims to minimize weight by optimizing the lengths of the forearm and upper arm links. SO1 focuses on maximizing manipulability by determining the optimal installation position of the manipulator. Meanwhile, SO2 optimizes the selection of motors and reducers that satisfy the required torque and angular velocity constraints, also aiming for weight reduction.

In this presentation, we optimize the entire manipulator using MDO by incorporating Directional Manipulability Index (DMI) and Directional Energy Index (DEI) into the installation position optimization within SO1.

Figure 1. Architecture of Multidisciplinary Design Optimization (MDO)

2 指標 / Indices

本稿では,従来用いていた可操作度[1]に代わる指標として,DMI[2]とDEI[3]を設置位置最適化に用いる.

前報では,経路方向に媒介変数$s$を導入し,経路方向の可操作性を評価する新たな指標DMI(Directional Manipulability Index) $\hat{w}_m$ を提案した.

In this study, we employ DMI (Directional Manipulability Index) [2] and DEI (Directional Energy Index) [3] for installation position optimization as alternative metrics to the conventional manipulability measure [1].

In our previous work, we introduced a new metric, DMI, by incorporating a trajectory parameter $s$ to evaluate manipulability in the trajectory direction. The DMI is defined as follows:

\[\hat{w}_m = \sqrt{ {\boldsymbol{J}_r}^+ \boldsymbol{J}\boldsymbol{J}^T { {\boldsymbol{J}_r}^+}^T }\]

ここで,$\boldsymbol{J}$は手先空間と関節空間をつなぐヤコビ行列であり,$\boldsymbol{J}_r$ は,手先空間と媒介変数空間をつなぐヤコビ行列である. したがって,DMIは媒介変数空間での可操作度を評価するものであり,手先空間での可操作性楕円体の経路方向の弦の長さに一致する.

また,DMIは消費エネルギーを考慮できない問題点があったため,経路方向を考慮したエネルギー指標としてDEI(Directional Energy Index) $\hat{w}_e$ を提案した.

where $\boldsymbol{J}$ is the Jacobian matrix that maps between the task space and the joint space, and $\boldsymbol{J}_r$ is the Jacobian matrix that maps between the task space and the parameterized trajectory space.

Thus, DMI evaluates manipulability within the parameterized trajectory space and corresponds to the chord length of the manipulability ellipsoid in the trajectory direction within the task space.

However, a limitation of DMI is its inability to account for energy consumption. To address this issue, we propose DEI (Directional Energy Index) as an energy-aware metric in the trajectory direction, defined as:

\[\hat{w}_e = \sqrt{ {\boldsymbol{J}_{jr}}^T \boldsymbol{J}^T \boldsymbol{M} \boldsymbol{J} {\boldsymbol{J}_{jr}} }\]

ここで,$\boldsymbol{J}_{jr}$は関節空間と媒介変数空間をつなぐヤコビ行列であり,$\boldsymbol{M}$は慣性行列である.これは,マニピュレータの媒介変数空間での消費エネルギーを評価するものである.

where $\boldsymbol{J}_{jr}$ is the Jacobian matrix mapping between the joint space and the parameterized trajectory space, and $\boldsymbol{M}$ is the inertia matrix. DEI quantifies the energy consumption within the parameterized trajectory space of the manipulator.


検証結果 – Result

DMIを用いて,2リンクマニピュレータを対象としたMDOの結果を示す.このとき,比較として従来用いていた可操作度と経路方向を考慮したエネルギー指標DEI(Directional Energy Index)[3]を用いた結果を示す. 本稿での設置位置最適化の最小化問題を以下に示す.

We present the MDO results for a two-link manipulator using DMI. For comparison, we also provide results obtained using the conventional manipulability measure and the Directional Energy Index (DEI) [3], which considers the trajectory direction.

The minimization problem for installation position optimization in this study is formulated as follows:

\[\mathrm{minimize}_{(x,y)} ~~ N\sum^N_{i=1} \frac{1}{\hat{w}_{m,i}}\] \[\mathrm{subject~to} ~~ x \in [x_\mathrm{min}, x_\mathrm{max}], y\in[y_\mathrm{min},y_\mathrm{max}]\]

$(x,y)$ はグローバル座標系におけるマニピュレータのベース位置の絶対座標,$N\in\mathbb{Z}$ は目標経路のデータ数,$\hat{w}_{m,i}$ は経路上の $i$ 番目の点における可操作性指標である.$(x,y)$ を設計変数として目標経路の各データ点での可操作性指標の平均値の逆数を最小化し,可操作性指標の平均値が最大となる$(x,y)$ を得る.

表1に2リンクマニピュレータを対象として線分軌道を与えたときにMDOによってリンクの長さ,マニピュレータの設置位置,モータ・減速機を最適化した結果を示す.左図は設置位置最適化の指標に可操作度を指標に用いたとき,中央は提案指標DMIを用いたとき,右図は提案指標DEIを用いたときの結果である.可操作度は,可操作性楕円体が最大となるように設置位置やリンクの長さが最適化されているのに対して,DMIでは軌道方向と可操作性楕円体の長軸と一致するように最適化されている.また,可操作度を用いて最適化した結果と比較して,DMIやDEIでは設置位置と軌道が近い位置に最適化された影響で,リンクの長さが短くなり,結果として軽量化したと考えられる.

where $(x,y)$ represents the absolute coordinates of the manipulator’s base position in the global coordinate system, $N \in \mathbb{Z}$ is the number of data points along the target trajectory, and $\hat{w}_{m,i}$ denotes the manipulability index at the $i^\mathrm{th}$ point on the trajectory. The optimization aims to minimize the reciprocal of the average manipulability index over all trajectory points, thereby determining $(x,y)$ that maximizes the average manipulability.

Table 1 presents the optimization results for a two-link manipulator following a linear trajectory. The table shows the optimized link lengths, installation position, motor selection, and reduction ratio obtained through MDO. The left figure corresponds to results using the conventional manipulability measure, the center figure shows results using the proposed DMI, and the right figure presents results using the proposed DEI.

When optimized using the conventional manipulability measure, the installation position and link lengths are adjusted to maximize the manipulability ellipsoid. In contrast, DMI-based optimization aligns the major axis of the manipulability ellipsoid with the trajectory direction. Additionally, compared to optimization using the conventional manipulability measure, DMI and DEI-based optimization resulted in shorter link lengths. This outcome suggests that the manipulator was positioned closer to the trajectory, leading to a lighter overall system.

Table 1. Multidisciplinary Design Optimization for a Two-Link Manipulator

  Manipulability DMI DEI
Manipulator itself
and its Position
Length of links[m]
(Link 1 / Link 2)
0.878 / 0.954 0.289 / 0.620 0.272 / 0.382
Motor output [W]
(Joint 1 / Joint 2)
50 / 50 50 / 50 100 / 50
Reduction ratio
(Joint 1 / Joint 2)
1 / 1 1 / 1 1 / 1
Total mass [kg] 41.46 21.56 14.61

結論 – Conclusion

本稿では,2リンクマニピュレータを対象として複合領域設計最適化を用いてリンクの長さ,マニピュレータの設置位置,モータと減速機を最適化した.このとき,前報までに提案したDMI(Directional Manipulability Index)とDEI(Directional Energy Index)を設置位置最適化の指標に用いた.この結果,従来用いていた可操作度と比較して,DMIとDEIを用いることによって軽量化できることが確認できた.

In this study, we applied Multidisciplinary Design Optimization (MDO) to a two-link manipulator to optimize its link lengths, installation position, motor selection, and reduction ratio. For installation position optimization, we employed the Directional Manipulability Index (DMI) and Directional Energy Index (DEI), which were proposed in our previous work. The results confirmed that, compared to the conventional manipulability measure, using DMI and DEIled to a lighter system design.


This work was supported by JSPS KAKENHI Grant Number 23K03755.


参考文献 - References

[1] 吉川恒夫,“ロボットアームの可操作度”, RSJ誌, vol.2, No.1, pp.63–67, 1984.

[2] 日比野圭歩, 遠藤央, Shan Zexin, 筒井幸雄, “可操作性楕円体の異方性を考慮した指標に基づく2 リンクマニピュレータを対象とした軌道最適化”, RSJ2024, 1I4-04, 2024.

[3] Kaho Hibino, Mitsuru Endo, Zexin Shan, Yukio Tsutsui, “Location Optimization of Manipulator to Minimize Energy Considering the Path Direction”, SII2025, ThuP1T2.7, 2025.


The content on this page is not freely available for reproduction or redistribution. Unauthorized use may lead to legal consequences.