JP2004195554A - Leg type mobile robot - Google Patents

Abstract

<P>PROBLEM TO BE SOLVED: To perform stable and highly efficient operation by dynamically or statically controlling an actuator characteristic when moving the whole machine body in the middle of overturning/falling. <P>SOLUTION: An actuator of respective joint parts is increased in attitudinal stability by performing highly accurate positioning control on these respective joint parts by maintaining the actuator characteristic of the joint parts related to stable area control with respective stages of overturning operation so that low area gain increases, a phase-lead quantity reduces, and viscous resistance of a joint increases. Thus, operational stability can be increased by securing joint positioning accuracy being a main component for controlling a quantity of ΔS/ΔT being a norm when controlling the overturning operation of the machine body. <P>COPYRIGHT: (C)2004,JPO&NCIPI

Description

【００６８】
また、モータと減速器の伝達関数表現モデルＧ（ｓ）は以下の式で表される。但し、Ｋはモータ・ゲインであり、Ｊはモータの慣性モーメントであり、Ｄはモータの粘性抵抗係数である。
【００６９】
【数２】

【００７０】
まず、図５に示したサーボ制御器において、位相補償型制御の設計例として位相補償帯域を任意に選択する例（位相補償量が一定で周波数帯域を任意に選ぶ）について、図７を参照しながら説明する。同図において、
【００７１】
▲１▼Ｃ（ｓ）−１：１．０〜１００Ｈｚの帯域で約＋５．６ｄＢのゲイン増幅、約＋１８ｄｅｇの位相進みを与えている。
▲２▼Ｃ（ｓ）−２：０．１〜１０Ｈｚの帯域で約＋５．６ｄＢのゲイン増幅、約＋１８ｄｅｇの位相進みを与えている。
▲３▼Ｃ（ｓ）−３：１０〜１ｋＨｚの帯域で約＋５．６ｄＢのゲイン増幅、約＋１８ｄｅｇの位相進みを与えている。
【００７２】
このように、位相補償を施す周波数帯域を任意に選択することで、アクチュエータの周波数特性を自在に設定することができる。したがって、このようなアクチュエータで構成されるロボットの関節軸は、機体の姿勢や動作の局面に応じて、周波数特性を動的に調整することができる。
【００７３】
なお、図７に示す例では、位相進み補償の例を示したが、位相遅れ補償場合も同様に任意の周波数帯域で任意の位相遅れ量を設定することができる。
【００７４】
次いで、図５に示したサーボ制御器において、位相補償型制御の設計例として位相補償を施す量を任意に選択する例（周波数帯域が一定で位相補償量を任意に選ぶ）について、図８を参照しながら説明する。同図において、
【００７５】
▲４▼Ｃ（ｓ）−４：４．０〜７０Ｈｚの帯域で約＋３．５ｄＢのゲイン増幅、約＋１２ｄｅｇの位相進みを与えている。
▲５▼Ｃ（ｓ）−５：２．０〜７０Ｈｚの帯域で約＋５．６ｄＢのゲイン増幅、約＋１８ｄｅｇの位相進みを与えている。
▲６▼Ｃ（ｓ）−６：１．０〜７０Ｈｚの帯域で約＋６．５ｄＢのゲイン増幅、約＋２１ｄｅｇの位相進みを与えている。
【００７６】
このように、位相補償を施す量を任意に選択することで、アクチュエータの周波数特性を自在に設定することができる。したがって、このようなアクチュエータで構成されるロボットの関節軸は、機体の姿勢や動作の局面に応じて、周波数特性を動的に調整することができる。
【００７７】
なお、図８に示す例では、位相進み補償の例を示したが、位相遅れ補償の場合も同様に任意の周波数帯域で任意の位相遅れ量を設定することができる。
【００７８】
次いで、図５に示したサーボ制御器において、Ｋで示した直列補償ゲインの大小を変更する制御器の設計例について、図９を参照しながら説明する。同図は、図９において、Ｋを±３ｄＢだけ上下させたことに相当する。図示の通り、直列補償ゲインの大小も任意に設定することができる。
【００７９】
図７〜図９に示した内容をロボットの関節軸駆動用のアクチュエータに適用するためには、これらの制御器を構成するパラメータを動的又は静的に変更するための通信プロトコルを実装する。これによって、ロボットの各関節軸にさまざまな特性を与えることができる。
【００８０】
次いで、これらの特性を備えたアクチュエータのサーボ制御器を実装したときのアクチュエータの特性について説明する。
【００８１】
図１０には、図７に示したように位相補償量が一定で周波数帯域を任意に選ぶようにアクチュエータのサーボ制御器を実装したときの開ループ特性を示している。
【００８２】
▲１▼Ｃ（ｓ）−１：１．０〜１００Ｈｚの帯域で約＋５．６ｄＢのゲイン増幅、約＋１８ｄｅｇの位相進みを与える
→ 全体的にゲインが高めとなるので、位置決め精度と追従性がよくなるが、エネルギーロスとなり易い。また、負荷が大きくなると不安定になる可能性がある。
▲２▼Ｃ（ｓ）−２：０．１〜１０Ｈｚの帯域で約＋５．６ｄＢのゲイン増幅、約＋１８ｄｅｇの位相進みを与える
→ Ｃ（ｓ）−１とＣ（ｓ）−２の中間的な特性を持つ。
▲３▼Ｃ（ｓ）−３：１０〜１ｋＨｚの帯域で約＋５．６ｄＢのゲイン増幅、約＋１８ｄｅｇの位相進みを与える
→ 高域でのみ位相進み補償をしていることになるので、ゆっくりした動作時にはあまり効果が見られないが、走る、飛ぶ、踊るといった速い動作には効果がある。
【００８３】
このように、位相補償を施す周波数帯域を任意に選択することで、アクチュエータの周波数特性を自在に設定することができる。したがって、このようなアクチュエータで構成されるロボットの関節軸は、機体の姿勢や動作の局面に応じて、周波数特性を動的に調整することができる。
【００８４】
また、図１１には、図１０に示したうち高域でのみ位相進み補償を施した例Ｃ（ｓ）−３においてさらに直列補償ゲインの制御を採り入れた様子を示している。この場合、図５に示した例と同様に、同位相においてゲインが上下する。
【００８５】
図１０に示した例では、位相補償例Ｃ（ｓ）−３ではゆっくりした動作時にはあまり効果が見られないが、図１１に示すように、低周波数帯域においてゲインを増やすことにより、低周波数帯域での制御偏差を少なくすることができる。この結果、ゆっくりした動作時でも指令値に対して少ない遅れで応答することが可能となる。
【００８６】
以上、アクチュエータのサーボ制御器においてサーボ制御器の比例ゲインだけでなく位相補償要素も関節各部位で調整するメカニズムについて説明してきた。これによって、安定且つ高効率な動作を実現するには局所的に力を増減させることや関節各部位のコンプライアンス（機械的受動性）を増減させることが可能となる。
【００８７】
例えば、位置制御系として関節各軸の動作を捉えたときは高ゲイン且つ高帯域なサーボ制御器を使って、制御偏差が少なくなるように制御した方がよいが、力学モデルとして捉えたときはポテンシャル・エネルギーや運動エネルギの作用を考慮してゲインを低くしたり位相補償している周波数帯を上下させたりすることも同時に行なうとよい。
【００８８】
Ｃ−２．アクチュエータ・モータの粘性抵抗
上記のＣ−１項で説明したようなアクチュエータの動作速度や動作時の機械的受動性という特性に加え、アクチュエータの粘性抵抗を可変に制御するという方式を採り入れることができる。
【００８９】
例えば、コイルへの供給電流を制御して所定の磁束分布を形成することにより回転トルクを発生させるタイプのモータは、一般に、コイル端子を電源電圧に接続する第１のトランジスタ・スイッチ群と、コイル端子を接地する第２のトランジスタ・スイッチ群からなるスイッチング動作回路をＰＷＭ制御により駆動することによって、コイル電流を制御し、所望のトルク又は回転位置、回転速度などを得るようになっている。
【００９０】
ここで、モータ・コイルが非通電の期間においてオープン状態となるタイミングにおいて、モータ・コイルに通電された電流（厳密には電荷）が抜けてしまうため、トルクのロスになる。またコギングによるトルクむらの影響を受け易くなる。
【００９１】
このような場合、モータ・コイルが非通電の期間においても、コイルがオープン状態とならない短絡（ショート）状態を形成することにより、モータ・コイルに通電された電流（厳密には電荷）が抜けないようにすることができる。このとき、モータのコイルには、永久磁石側からの磁束密度により逆起電力が発生する。この逆起電力により、モータの回転方向の逆方向に力が作用するので、外力による回転に対する粘性抵抗を作り出すことができ、ブレーキに類似した効果を得ることができる。このようなモータへの粘性抵抗により、トルク・ロスがなくコギングによるトルクむらの影響が軽減される。
【００９２】
一方、モータの非通電時において、このようなコイルの短絡状態を形成した場合、上述したように、モータに一種の粘性抵抗を与えることができるが、このようなモータをロボットに使用した場合、コイル・ショートによるブレーキの影響のため、コンプライアンス（機械的受動性）がなくなってしまうという問題を招来する。
【００９３】
そこで、モータ・コイルの非通電時におけるコイルのオープン状態と短絡状態の期間の比率を、所望の機械的特性に応じて調整することにより、モータ・コイルがオープン状態となるタイミングにおけるモータ・コイルに通電された電流（厳密には電荷）が抜けてしまうことによるトルクのロスやコギングによるトルクむらの問題と、モータ・コイルの非通電時におけるコイル・ショートによるブレーキの影響のためコンプライアンス（機械的受動性）がなくなるという問題を、ともに解決することができる。
【００９４】
ここで、モータ・コイルの通電及び非通電状態の比率はＰＷＭ制御により実現することができるが、モータ・コイルの非通電状態におけるコイルのオープン状態と短絡状態の期間の比率も同様にＰＷＭ制御を用いて実現することができる。
【００９５】
図１２には、コイル電流の制御メカニズムを適用したＤＣモータのコイル電流供給用の電流制御回路の等価回路の構成例を示している。
【００９６】
同図に示す電流制御回路は、フルブリッジ構成であり、ｐｎｐ型のトランジスタＡ'とｎｐｎ型のトランジスタＡを順方向接続した回路と、同じくｐｎｐ型のトランジスタＢ'とｎｐｎ型のトランジスタＢを順方向接続した回路を電源電圧Ｖ_ccとグランドＧＮＤの間に並列接続し、さらにトランジスタＡ'及びＡの中間点とトランジスタＢ'及びＢの中間点を固定子の単相コイルで接続している。
【００９７】
トランジスタＡ'及びＢをオンにするとともに、トランジスタＡ及びＢ'をオフにすることによって、モータ・コイルには、図示の矢印方向の電流Ｉ_ｍが流れる。また、トランジスタＡ'及びＢをオフにすることによって、コイルはオープン状態となって、電流Ｉ_ｍは流れなくなる。また、トランジスタＡ'及びＢをオフにするとともに、トランジスタＡ'及びＢ'をオンにすることによって、モータ・コイルは短絡（ショート）状態となる。
【００９８】
ＰＷＭ制御論理回路は、図示しない中央制御部からの電流軸電流指令（又はトルク指令）に基づいてコイルへの電流指令を生成し、これら電流指令に基づいて各トランジスタをＰＷＭ方式にてスイッチング制御する。すなわち、トランジスタＡ'及びＢをオンにするとともにトランジスタＡ及びＢ'をオフにしてコイル電流Ｉ_ｍを流す通電期間と、トランジスタＡ'及びＢをオフにしてコイルを非通電にする非通電期間を交互に生成する。
【００９９】
付加論理回路は、さらにＰＷＭ制御論理回路が出力する信号Ａ₀及びＡ₀'、並びＢ₀及びＢ₀'をオン／オフ動作を制御する制御論理を付加論理により切り替える。これによって、モータ・コイルの非通電状態におけるコイルのオープン状態と短絡状態の切替動作を行なう。図１３には、付加論理回路の具体的な回路構成を示している。
【０１００】
ＰＷＭ制御論理回路からのトランジスタＡ'制御用の信号Ａ₀'とトランジスタＢ₀'制御用の信号Ｂ₀'の論理積と、トランジスタＡ制御用の信号Ａ₀とトランジスタＢ制御用の信号Ｂ₀の排他的論理和がとられ、さらにこれらの論理演算値の論理積を反転したものがＢＲＡＫＥ＿ＰＷＭ制御信号と論理和される。この論理和の結果を元の各トランジスタ制御用信号との間で論理積を取ったものが、最終的なそれぞれのトランジスタ制御用信号となる。
【０１０１】
付加論理回路は、ＢＲＡＫＥ＿ＰＷＭ制御信号がハイ・レベルが入力されると、コイル非通電時にコイルを短絡させるようにトランジスタ制御用信号を切り替える。通常のコイル非通電時において、ＰＷＭ制御論理回路からは、制御信号Ａ₀'及びＡ₀をハイ、Ｂ₀'及びＢ₀をローにするトランジスタ制御用信号が出力される。これに対し、付加論理回路は、ハイ・レベルのＢＲＡＫＥ＿ＰＷＭ制御信号が入力されると、ハイ状態のＡ₁'及びＡ₁をローに転じて、コイルの短絡状態を形成する。
【０１０２】
一方、付加論理回路は、ＢＲＡＫＥ＿ＰＷＭ制御信号がロー状態のときには、コイル非通電時において、ＰＷＭ制御論理回路からのトランジスタ制御用信号をそのまま出力するので、非通電時におけるコイルはオープン状態となる。
【０１０３】
図１４には、ＰＷＭ制御により所定のデューティ比を持つＢＲＡＫＥ＿ＰＷＭ制御信号が入力されたときの付加論理回路の各トランジスタ制御用信号の出力特性を、コイル電流波形特性及びトルク出力特性とともに示している。
【０１０４】
コイル非通電時にコイルを短絡状態にすると、過渡応答により、コイル電流がゼロに戻るまでの時間が長くなるが、コイルをオープン状態にするとその時間は短くなる。コイル非通電時の過渡応答特性は、ＢＲＡＫＥ＿ＰＷＭ制御信号のデューティ比に従って、これらの特性が混合されたものとなる。
【０１０５】
したがって、同図に示すように、コイル通電とコイル短絡というスイッチング動作を繰り返した場合、コイル非通電時にコイル電流がゼロに戻る前に次の通電が開始される。このときのコイルの最大電流は、コイル通電及び非通電の動作の度に逐次上昇していくが、その上昇傾向はデューティ比すなわちＢＲＡＫＥ＿ＰＷＭ制御信号がハイ・レベルとなる比率にほぼ比例する。同様に、コイル電流の実効値は図示の通り徐々に増加していくことになるが、その上昇傾向はデューティ比すなわちＢＲＡＫＥ＿ＰＷＭ制御信号がハイ・レベルとなる比率にほぼ比例する。
【０１０６】
また、モータの出力トルクＴは、コイル電流にモータのトルク定数Ｋ_ｔを乗じた値となるので（Ｔ＝Ｋ_ｔ・Ｉ）、同図からも判るように、コイル通電及び非通電を繰り返したとき、コイル電流の増大に伴って、モータ・トルクの実効値は増大していく。このときの上昇傾向はＢＲＡＫＥ＿ＰＷＭ制御信号のデューティ比すなわち同制御信号がハイ・レベルとなる比率にほぼ比例する。モータ・トルクの出力が上昇していく特性が、モータの粘性係数に相当する。言い換えれば、ＢＲＡＫＥ＿ＰＷＭ制御信号のデューティ比によってモータの粘性抵抗を動的に制御することが可能である。
【０１０７】
このようにＰＷＭ制御論理回路が付加論理回路へ供給するＢＲＡＫＥ＿ＰＷＭ制御信号のデューティ比をＰＷＭ制御することにより、モータ・コイルの非通電時におけるコイルのオープン状態と短絡状態の期間の比率を、所望の機械的特性に応じて調整することができる。
【０１０８】
したがって、モータ・コイルがオープン状態となるタイミングにおけるモータ・コイルに通電された電流（厳密には電荷）が抜けてしまうことによるトルクのロスやコギングによるトルクむらの問題と、モータ・コイルの非通電時におけるコイル・ショートによるブレーキの影響のためコンプライアンス（機械的受動性）がなくなるという問題を、ともに解決することができる。
【０１０９】
なお、上記の説明ではＤＣモータを例にとって説明したが、３相モータやその他のコイルへの供給電流を制御して所定の磁束分布を形成することにより回転トルクを発生させるタイプのモータにおいても同様に、非通電時のモータ・コイルを間歇的にオープン状態及び短絡状態に切り替えることによって、モータの所望の粘性抵抗を得ることができる。
【０１１０】
Ｃ−３．脚式移動ロボットへの適用
次いで、本実施形態に係るアクチュエータのサーボ制御器の特性制御並びにアクチュエータ自体の特性制御のメカニズムを各関節部位に適用した２足歩行の脚式移動ロボットについて説明する。
【０１１１】
上述したように、アクチュエータのサーボ制御器において、比例ゲインと位相補償要素を調整することにより、ロボットの関節各部位において必要な位置決め精度、機械的受動性（コンプライアンス）、動作速度を任意に設定することができる。また、アクチュエータ・モータのコイルへの非通電時においてコイルを間歇的に短絡状態又はオープン状態に切り替えることによって、モータの粘性抵抗を調整し振動などの外乱に対するロバスト性を変更することができる。
【０１１２】
さらに、これらアクチュエータのサーボ制御器におけるゲイン及び位相補償制御とアクチュエータ・モータの粘性抵抗の制御とを組み合わせることによって、位置決め精度が重要な部分に適用することができるアクチュエータの周波数特性、あるいは速応性とコンプライアンスが重要となる部分に適用することができるアクチュエータの周波数特性を得ることができる。
【０１１３】
図１５には、位置決め精度が重要な部分に適用することができるアクチュエータの周波数特性を示している。
【０１１４】
この場合、サーボ制御器の比例ゲインを大きくとって、系全体のゲインを大きくし、低周波数帯域までゲインを取れるようにする。また、周波数特性を同図に示すように高域で位相進み量が小さくなるようにし、速応性にはあまり寄与しないが安定性を保証できるようにする。また、モータの粘性抵抗を大きくし、振動などの外乱にもロバストになるようにする。要言すれば、図示の特性は、位置決め精度を優先して振動などの外乱にもロバストな特性である。
【０１１５】
また、図１６には、速応性とコンプライアンスが重要となる部分に適用することができるアクチュエータの周波数特性を示している。
【０１１６】
この場合、比例ゲインを小さくして、系全体のゲインを小さくすることで、低周波数帯域のゲインを小さくし、機械的受動性（コンプライアンス）を得易くする。また、周波数特性を同図に示すように中高域で位相進み量が大きくなるようにし、速応性を得るようにする。また、モータの粘性抵抗を小さくし、機械的受動性（コンプライアンス）を得易いようにする。要言すれば、図示の特性は、機械的受動性（コンプライアンス）と速応性を優先した特性である。
【０１１７】
図１〜３に示した脚式移動ロボットの関節自由度は、上述したサーボ制御器を備えたアクチュエータによって実現される。各関節部位において使用されるアクチュエータのゲイン・位相補償特性の基本的な制御例について、以下に詳解する。
【０１１８】
（１）首部分に適用されるアクチュエータの特性
首部分では、位置決め精度を優先させるため、比例ゲインを高く設定する。また、動作速度を保持しつつ、比例ゲインを上げた分の安定性を損なわないように、位相進み量を少なく設定する。また、胴体より下の部分の動作時に発生する振動外乱に対してロバスト性を得るため、関節の粘性抵抗を大きく設定する。
【０１１９】
（２）肩・肘部分に適用されるアクチュエータの特性
歩行やダンスなどの連続動作を行なうときは、位置決め特性よりも機械的受動性が高くなるような特性をアクチュエータに与える。動作に受動性を持たせるため、関節の粘性抵抗を小さくする。また、動作に受動性を持たせるとともに、エネルギ消費量を減らすため、比例ゲインを低く設定する。また、動作速度を大きくするため、位相進み補償を行なう周波数帯域を高く、位相進み量を大きく設定する。動作によっては振り子のように往復運動を行なうだけのときもある。そのときは、関節の粘性抵抗と比例ゲインを最小にして機械的受動性（コンプライアンス）を得るようにし、力学的なエネルギを動作のために用い易くする。
【０１２０】
一方、物を押したり引っ張ったりするなどの力を使う動作を行なうときには、負荷トルク値により、位置決め精度優先の特性と、機械的受動性の特性を動的に入れ替えるように制御する。負荷トルク値に対してさらに力を発生させたいときは比例ゲインを高く、関節の粘性抵抗を大きくする。また、負荷トルク値に対して定負荷となるように倣わせる動作をする場合には、上位からの位置指令値による調整に加え、アクチュエータ内部トルク・センサで検知した負荷トルクに応じて比例ゲインを低く、関節の粘性抵抗を小さくして、機械的受動性（コンプライアンス）を得るようにする。
【０１２１】
（３）体幹部分に適用されるアクチュエータの特性
自分自身の動作による振動外乱にロバスト性を得るため、関節の粘性抵抗を大きくする。あるいは、位置決め精度を優先させるため、比例ゲインを高く設定する。あるいは、動作速度を保持しつつ比例ゲインを上げた分の安定性を損なわないように、位相進み量を少なく設定する。
【０１２２】
（４）股関節部分に適用されるアクチュエータの特性
自分自身の動作による振動外乱にロバスト性を得るため、関節の粘性抵抗を大きくする。あるいは、位置決め精度を優先させるため、比例ゲインを高く設定する。あるいは、動作速度を保持しつつ比例ゲインを上げた分の安定性を損なわないように、位相進み量を少なく設定する。
【０１２３】
（５）膝部分に適用されるアクチュエータの特性
遊脚時及び着床瞬間時においては、位置決め精度よりも機械的受動性が高くなるような特性に制御する。動作に受動性を持たせるため、関節の粘性抵抗を小さくする。また、動作に受動性を持たせるとともにエネルギ消費量を減らすため、比例ゲインを低く設定する。また、動作速度を大きくするため、位相進み補償を行なう周波数領域を高く、位相進み量を大きく設定する。
【０１２４】
一方、支持脚時においては、機械的受動性よりも位置決め精度が高くなるような特性に制御する。自分自身の動作による振動外乱にロバスト性を得るため、関節の粘性抵抗を大きくする。あるいは、位置決め精度を優先させるため、比例ゲインを高く設定する。あるいは、動作速度を保持しつつ比例ゲインを上げた分の安定性を損なわないように、位相進み量を少なく設定する。
【０１２５】
（６）足首部分に適用されるアクチュエータの特性
遊脚時及び着床瞬間時においては、位置決め精度よりも機械的受動性が高くなるような特性に制御する。足首部分着床による衝撃を緩和するため、関節粘性を小さく設定し、機械的受動性（コンプライアンス）を得るようにする。また、足首部分着床による衝撃を緩和するため、比例ゲインを低く設定し、機械的受動性（コンプライアンス）を得るようにする。また、動作速度を大きくとるため、位相補償を行なう周波数領域を高く、位相進み量を大きく設定する。
【０１２６】
一方、支持脚時において、足首部分の発生トルクを大きくとり、且つ自分自身の動作による振動外乱にロバスト性を得るため、関節の粘性抵抗を大きくとる。また、足首部分の位置決め精度を向上させるため、比例ゲインを高く設定する。また、動作速度を保持しつつ、比例ゲインを上げた分の安定性を損なわないように位相進み量を少なく設定する。
【０１２７】
Ｄ．脚式移動ロボットの姿勢安定制御
次いで、本実施形態に係る脚式移動ロボット１００における、脚式作業時における姿勢安定化処理、すなわち足部、腰、体幹、下肢運動などからなる全身協調運動実行時における姿勢の安定化処理の手順について説明する。
【０１２８】
姿勢安定制御には、ＺＭＰを姿勢安定制御に用いる。本実施形態では、ロボットの機体上の制御目標点として質量操作量が最大となる部位、例えば腰部をローカル座標原点に設定する。そして、この制御目標点に加速度センサなどの計測手段を配置して、その位置における姿勢や加速度を直接計測して、ＺＭＰに基づく姿勢安定制御を行なう。さらに路面との接触部位である足部に加速度センサを配備することにより、制御に用いるローカル座標とその座標を直接的に計測して、ＺＭＰ位置に最も近い足部で直接ＺＭＰ方程式を組み立てる。
【０１２９】
Ｄ−１．ＺＭＰ方程式の導入
本実施形態に係る脚式移動ロボット１００は、無限すなわち連続的な質点の集合体である。但し、ここでは有限数で離散的な質点からなる近似モデルに置き換えることによって、安定化処理のための計算量を削減するようにしている。より具体的には物理的には図３に示す多関節自由度構成を具備する脚式移動ロボット１００を、図１７に示すように多質点近似モデルに置き換えて取り扱う。図示の近似モデルは、線形且つ非干渉の多質点近似モデルである。
【０１３０】
図１０において、Ｏ−ＸＹＺ座標系は絶対座標系におけるロール、ピッチ、ヨー各軸を表し、また、Ｏ'−Ｘ'Ｙ'Ｚ'座標系はロボット１００とともに動く運動座標系におけるロール、ピッチ、ヨー各軸を表している。但し、図中におけるパラメータの意味は以下の通りである。また、ダッシュ（´）付きの記号は運動座標系を記述するものと理解されたい。
【０１３１】
【数３】

【０１３２】
図１７に示す多質点モデルでは、ｉはｉ番目に与えられた質点を表す添え字であり、ｍ_iはｉ番目の質点の質量、ｒ_i'はｉ番目の質点の位置ベクトル（但し運動座標系）を表すものとする。本実施形態に係る脚式移動ロボット１００の機体重心は腰部付近に存在する。すなわち、腰部は、質量操作量が最大となる質点であり、図１７では、その質量はｍh、その位置ベクトル（但し運動座標系）はｒ'_h（ｒ'_hx，ｒ'_hy，ｒ'_hz）とする。また、機体のＺＭＰの位置ベクトル（但し運動座標系）をｒ'_zmp（ｒ'_zmpx，ｒ'_zmpy，ｒ'_zmpz）とする。なお、腰部情報における各回転角（θ_hx，θ_hy，θ_hz）は、脚式移動ロボット１００における腰部の姿勢すなわちロール、ピッチ、ヨー軸の回転を規定する。
【０１３３】
世界座標系Ｏ−ＸＹＺは絶対座標系であり、不変である。本実施形態に係る脚式移動ロボット１００は、腰部と両脚の足部にそれぞれ加速度センサ９３、９４、９６が配置されており、これらセンサ出力により腰部並びに立脚それぞれと世界座標系の相対位置ベクトルｒ_qが直接検出される。これに対し、運動座標系すなわち機体のローカル座標系はＯ−Ｘ’Ｙ’Ｚ’は、ロボットともに動く。
【０１３４】
機体のＺＭＰ方程式は、制御目標点において印加される各モーメントの釣合い関係を記述したものである。図５に示したように、機体を多数の質点ｍ_iで表わし、これらを制御目標点とした場合、すべての制御目標点ｍ_iにおいて印加されるモーメントの総和を求める式がＺＭＰ方程式である。
【０１３５】
世界座標系（Ｏ−ＸＹＺ）で記述された機体のＺＭＰ方程式、並びに機体のローカル座標系（Ｏ−Ｘ’Ｙ’Ｚ’）はそれぞれ以下の通りとなる。
【０１３６】
【数４】

【０１３７】
上式は、各質点ｍ_iにおいて印加された加速度成分により生成されるＺＭＰ回り（半径ｒ_i−ｒ_zmp）のモーメントの総和と、各質点ｍiに印加された外力モーメントＭ_iの総和と、外力Ｆ_kにより生成されるＺＭＰ回り（ｋ番目の外力Ｆ_kの作用点をｓ_kとする）のモーメントの総和が釣り合うということを記述している。
【０１３８】
このＺＭＰ釣合い方程式は、総モーメント補償量すなわちモーメント・エラー成分Ｔを含んでいる。このモーメント・エラーをゼロ又は所定の許容範囲内に抑えることによって、機体の姿勢安定性が維持される。言い換えれば、モーメント・エラーをゼロ又は許容値以下となるように機体運動（足部運動や上半身の各部位の軌道）を修正することが、ＺＭＰを安定度判別規範とした姿勢安定制御の本質である。
【０１３９】
本実施形態では、腰部と左右の足部にそれぞれ加速度センサ９６，９３及び９４が配設されているので、これらの制御目標点における加速度計測結果を用いて直接的に且つ高精度に上記のＺＭＰ釣合い方程式を導出することができる。この結果、高速でより厳密な姿勢安定制御を実現することができる。
【０１４０】
Ｄ−２．全身協調型の姿勢安定制御
図１８には、脚式移動ロボット１００において、ＺＭＰを安定度判別規範に用いて安定歩行可能な機体運動を生成するための処理手順をフローチャートの形式で示している。但し、以下の説明では、図１７に示すような線形・非干渉多質点近似モデルを用いて脚式移動ロボット１００の各関節位置や動作を記述するものとする。
【０１４１】
まず、足部運動の設定を行う（ステップＳ１）。足部運動は、２以上の機体のポーズを時系列的に連結されてなるモーション・データである。モーション・データは、例えば、足部の各関節角の変位を表わした関節空間情報と、関節位置を表わしたデカルト空間情報で構成される。
【０１４２】
次いで、設定された足部運動を基にＺＭＰ安定領域を算出する（ステップＳ２）。ＺＭＰは、機体に印加されるモーメントがゼロとなる点であり、基本的には足底接地点と路面の形成する支持多角形の辺上あるいはその内側に存在する。ＺＭＰ安定領域は、この支持多角形のさらに内側に設定された領域であり、該領域にＺＭＰを収容させることによって機体を高度に安定した状態にすることができる。
【０１４３】
そして、足部運動とＺＭＰ安定領域を基に、足部運動中におけるＺＭＰ軌道を設定する（ステップＳ３）。
【０１４４】
また、機体の上半身（股関節より上側）の各部位については、腰部、体幹部、上肢、頭部などのようにグループ設定する（ステップＳ１１）。そして、各部位グループ毎に希望軌道を設定する（ステップＳ１２）。
【０１４５】
次いで、各部位のグループ設定の調整（再グルーピング）を行ない（ステップＳ１３）、さらにこれらグループに対して優先順位を与える（ステップＳ１４）。ここで言う優先順位とは、機体の姿勢安定制御を行なうための処理演算に投入する順位のことであり、例えば質量操作量に応じて割り振られる。この結果、機体上半身についての各部位についての優先順位付き希望軌道群が出来上がる。
【０１４６】
また、機体上半身の各部位グループ毎に、モーメント補償に利用できる質量を算出しておく（ステップＳ１５）。
【０１４７】
そして、足部運動とＺＭＰ軌道、並びに上半身の各部位グループ毎の希望起動群を基に、ステップＳ１４により設定された優先順位に従って、各部位グループの運動パターンを姿勢安定化処理に投入する。
【０１４８】
この姿勢安定化処理では、まず、処理変数ｉに初期値１を代入する（ステップＳ２０）。そして、優先順位が先頭からｉ番目までの部位グループについての目標軌道設定時における、目標ＺＭＰ上でのモーメント量すなわち総モーメント補償量を算出する（ステップＳ２１）。目標軌道が算出されていない部位については、希望軌道を用いる。
【０１４９】
次いで、ステップＳ１５において算出された当該部位のモーメント補償に利用できる質量を用いて、そのモーメント補償量を設定して（ステップＳ２２）、モーメント補償量を算出する（ステップＳ２３）。
【０１５０】
次いで、算出されたｉ番目の部位のモーメント補償量を用いて、ｉ番目の部位についてのＺＭＰ方程式を導出して（ステップＳ２４）、当該部位のモーメント補償運動を算出することにより（ステップＳ２５）、優先順位が先頭からｉ番目までの部位についての目標軌道を得ることができる。
【０１５１】
このような処理をすべての部位グループについて行なうことにより、安定運動（例えば歩行）が可能な全身運動パターンが生成される。
【０１５２】
腰部と左右の足部にそれぞれ加速度センサ９６，９３及び９４が配設されているので、これらの制御目標点における加速度計測結果を用いて直接的に且つ高精度に上記のＺＭＰ釣合い方程式を導出することができる。この結果、図１２に示すような処理手順に従ってＺＭＰ安定度判別規範に基づく姿勢安定制御を高速でより厳密に実行することができる。
【０１５３】
Ｅ．脚式移動ロボットの転倒オペレーション
前項Ｄで説明したように、本実施形態に係る脚式移動ロボット１００は、基本的には、ＺＭＰ安定度判別規範に基づいて、歩行時やその他の立脚作業時における姿勢安定制御を行ない、機体の転倒という事態の発生を最小限に抑えるようにしている。しかしながら、万一転倒を避けられなくなった場合には、機体へのダメージを極力防止するような動作パターンからなる転倒動作を行なうことにする。例えば、前述したＺＭＰ釣合い方程式において、過大な外力Ｆ又は外力モーメントＭが機体に印加された場合、機体動作のみによってモーメント・エラー成分Ｔをキャンセルことができなくなり、姿勢の安定性を維持できなくなる。
【０１５４】
図１９には、本実施形態に係る脚式移動ロボット１００における脚式作業中の機体の動作制御の概略的な処理手順をフローチャートの形式で示している。
【０１５５】
機体動作中は、左右の足部に配設した接地確認（床反力）センサ９１及び９２、加速度センサ９３及び９４、腰部に配設した加速度センサ９６のセンサ出力を用いて、ＺＭＰ釣合い方程式（前述）を立てて、足部の接地位置を常に計算する（ステップＳ３１）。
【０１５６】
例えば、機体に外力が印加されたとき、次の足部の接地位置を計画することができるかどうか、すなわち足部の行動計画によって外力によるモーメント・エラーを解消することができるかどうかを判別する（ステップＳ３２）。足部の設置位置を計画可能か否かは、脚部の各関節の可動角、各関節アクチュエータのトルク、関節力、角速度、角加速度などを考慮して判断する。勿論、外力が加わったときに、次の一歩だけでなく、数歩にまたがる脚式動作によりモーメント・エラーを解消するようにしてもよい。
【０１５７】
このとき、足部の計画が可能であれば、歩行やその他の脚式動作を継続する（ステップＳ３３）。
【０１５８】
他方、過大な外力又は外力モーメントが機体に印加されたために、足部の計画が不可能になった場合には、脚式移動ロボット１００は転倒動作を開始する（ステップＳ３４）。
【０１５９】
図１〜図２に示すような直立歩行型の脚式ロボットの場合、重心位置が高いことから、転倒時に不用意に床面に落下すると、ロボット自体、あるいは転倒により衝突する相手側にも致命的な損傷を与えてしまう危険がある。そこで、本実施形態では、転倒前に計画されている機体の軌道からＺＭＰ支持多角形が最小となるような姿勢に組み替えて、所定の転倒動作を実行する。基本的には、以下に示す２つの方針を基に転倒動作を探索していく。
【０１６０】
（１）機体の支持多角形の面積Ｓの時間ｔ当たりの変化量ΔＳ／Δｔを最小にする。
（２）床面落下時における支持多角形が最大となるようにする。
【０１６１】
このような転倒方法をとることにより、落下時に床面から受ける衝撃を全身に分散させることにより、ダメージを最小限に抑えることができる。
【０１６２】
図２０には、本実施形態に係る脚式移動ロボット１００が足部の計画不能により転倒動作を行なうための処理手順をフローチャートの形式で示している。転倒動作は、上述した基本方針に従って、高さ方向に連結された肩関節ピッチ軸４、体幹ピッチ軸９、股関節ピッチ軸１２、膝関節ピッチ軸１４を同期協調的に駆動させることによって実現される。このような処理手順は、実際には主制御部８１において所定の機体動作制御プログラムを実行して、各部を駆動制御することによって実現される。
【０１６３】
まず、機体の支持多角形の面積Ｓの時間ｔ当たりの変化量ΔＳ／Δｔを最小にするリンクを探索する（ステップＳ４１）。
【０１６４】
次いで、ステップＳ４１により選択されたリンクで変化量ΔＳ／Δｔを最小にする該リンクの目標着床点を探索する（ステップＳ４２）。
【０１６５】
次いで、先行ステップにより選択されたリンクを該目標着床点に着床することが、機体ハードウェアの制約上（各関節の可動角、各関節アクチュエータのトルク、関節力、角速度、角加速度など）、実行可能かどうかを判別する（ステップＳ４３）。
【０１６６】
先行ステップにより選択されたリンクを該目標着床点に着床することが不可能であると判定された場合には、時間の変化量Δｔを所定地だけ増分してから（ステップＳ４４）、ステップＳ４１に戻って、リンクの再選択、並びに該リンクの目標着床点の再設定を行なう。
【０１６７】
一方、先行ステップにより選択されたリンクを該目標着床点に着床することが可能である場合には、選択されたリンクを該目標着床点に着床する（ステップＳ４５）。
【０１６８】
次いで、機体の位置エネルギが最小かどうか、すなわち転倒動作が完了したかどうかを判別する（ステップＳ４６）。
【０１６９】
機体の位置エネルギがまだ最小ではない場合には、時間の変化量Δｔをさらに所定地だけ増分して（ステップＳ４７）、支持多角形を拡大するように次の目標着床点を設定する（ステップＳ４８）。
【０１７０】
次いで、選択されたリンクを該目標着床点に着床することが、機体ハードウェアの制約上（各関節の可動角、各関節アクチュエータのトルク、関節力、角速度、角加速度など）、実行可能かどうかを判別する（ステップＳ４９）。
【０１７１】
先行ステップにより選択されたリンクを該目標着床点に着床することが不可能であると判定された場合には、ステップＳ４１に戻って、リンクの再選択、並びに該リンクの目標着床点の再設定を行なう。
【０１７２】
一方、先行ステップにより選択されたリンクを該目標着床点に着床することが可能である場合には、ステップＳ４５に進んで、選択されたリンクを該目標着床点に着床する。
【０１７３】
そして、機体の位置エネルギが最小になると（ステップＳ４６）、機体の床面への着床が完了したことになるので、本処理ルーチン全体を終了する。
【０１７４】
図２１には、脚式移動ロボット１００を肩関節ピッチ軸４、体幹ピッチ軸９、股関節ピッチ軸１２、膝関節ピッチ軸１４などの高さ方向に連結された略平行な複数の関節軸からなるリンク構造体としてモデル化して、各関節ピッチ軸を同期強調的に駆動させて仰向け姿勢に向かって転倒していく動作を示している。
【０１７５】
ロボットは、リンク構造体のリンク端である足底のみで立位しているとする（図２１（１））。
【０１７６】
このとき、外力又は外力モーメントの印加により、ＺＭＰ釣合い方程式のモーメント・エラー項Ｔをキャンセルできなくなり、足底のみで形成するＺＭＰ安定領域の外にＺＭＰが逸脱したことに応答して、ＺＭＰを支持多角形に維持したまま転倒動作を開始する。
【０１７７】
転倒動作では、まず、機体の支持多角形の面積Ｓの時間ｔ当たりの変化量ΔＳ／Δｔを最小にするリンクを探索するとともに、手先を含むリンクで変化量ΔＳ／Δｔを最小にするような手先の目標着床点を探索する。そして、選択されたリンクを目標着床点に着床することが機体ハードウェアの制約上（各関節の可動角、各関節アクチュエータのトルク、関節力、角速度、角加速度など）、実行可能かどうかを判別する。
【０１７８】
機体ハードウェア上実行可能である場合には、既に着床している足底リンクに加え、他のリンクが着床する。そして、これら着床リンクで形成される最小支持多角形内にＺＭＰを移動する（図２１（２））。
【０１７９】
次いで、機体ハードウェアが許容する限り、着床点を移動して、支持多角形を拡大していく（図２１（３））。
【０１８０】
そして、各関節の可動角、各関節アクチュエータのトルク、関節力、角速度、角加速度など機体ハードウェアの制約から、もはや着床点を移動することができなくなると、今度は、着床中のリンクに挟まれている離床リンクを着床することができるかどうかを判別する。
【０１８１】
機体ハードウェア上、着床リンク間の離床リンクを着床することが可能である場合には、これらを着床して、着床リンク数を増やす（図２１（４））。
【０１８２】
さらに、機体ハードウェアが許容する限り、着床点を移動して、支持多角形を拡大していく（図２１（５））。
【０１８３】
そして最後に、高さ方向に連結された略平行な複数の関節軸からなるリンク構造体の一端側から１以上のリンクと、他端がわから２以上のリンクを離床させて、且つ、それらの中間に位置するリンクを１以上着床させ、さらに足部を着床させた状態で、ＺＭＰを支持多角形内に維持しながら、支持多角形が最大となる姿勢を形成する。この姿勢において、機体の位置エネルギが最小であれば、転倒動作は完了である。
【０１８４】
図２２には、実機が立位姿勢から仰向け姿勢に転倒していく様子を示している。同図に示す例では、機体の前方から外力を受けて後ろ向きに転倒動作を行なう様子を示しているが、Ａ〜Ｄの４段階に分けて動作状況を捉えることができる。
【０１８５】
（Ａ）転倒前の両脚支持の状態
姿勢制御されている状態であり、すべての関節部位のアクチュエータが図１５に示す周波数特性に設定されている。
【０１８６】
この場合、サーボ制御器の比例ゲインを大きくとって、系全体のゲインを大きくし、低周波数帯域までゲインを取れるようにする。また、周波数特性を同図に示すように高域で位相進み量が小さくなるようにし、速応性にはあまり寄与しないが安定性を保証できるようにする。また、モータの粘性抵抗を大きくし、振動などの外乱にもロバストになるようにする。この結果、機体全体として、位置決め精度を優先して振動などの外乱にもロバストとなる。
【０１８７】
（Ｂ）転倒に至る外力・衝撃を受け、衝撃を最小限にするための転倒動作を開始足首ロール軸、足首ピッチ軸、膝ピッチ軸など位置エネルギ制御に関わる関節部位のアクチュエータを、図１６に示す周波数特性に設定する。また、それ以外の安定領域制御に関わる関節部位のアクチュエータを、図１５に示す周波数特性に維持する。
【０１８８】
足首ロール軸、足首ピッチ軸、膝ピッチ軸など位置エネルギ制御に関わる関節部位では、速応性とコンプライアンスが重要となる。そこで、比例ゲインを小さくして、系全体のゲインを小さくすることで、低周波数帯域のゲインを小さくし、機械的受動性（コンプライアンス）を得易くする。また、周波数特性を同図に示すように中高域で位相進み量が大きくなるようにし、速応性を得るようにする。また、モータの粘性抵抗を小さくし、機械的受動性（コンプライアンス）を得易いようにする。
【０１８９】
（Ｃ）足底が離床し、足底面で請けていた床反力がゼロになる
股関節ピッチ軸など位置エネルギ制御に関わる関節部位のアクチュエータを、さらに図１６に示す周波数特性に設定する。この結果、当該関節部位では機械的受動性（コンプライアンス）と速応性を優先した特性が得られる。また、それ以外の安定領域制御に関わる関節部位のアクチュエータを、図１５に示す周波数特性に維持する。
【０１９０】
（Ｄ）ロボットが背面に着床し、転倒動作が終了
着床と同時に残りすべての関節部位のアクチュエータを図１６に示す周波数特性に設定する。この結果、機体全体として機械的受動性（コンプライアンス）と速応性を優先した特性が得られる。
【０１９１】
アクチュエータの特性を図１５に示すように低域ゲインを大きく、位相進み量を小さく、関節の粘性抵抗を大きくするように設定することで、高精度な位置決めが可能となる。したがって、（Ｂ）及び（Ｃ）の状態のときに、転倒動作を制御する際の規範となるΔＳ／Δｔの量を制御するための主成分となる関節の位置決め精度が確保され、動作の安定性が増す。
【０１９２】
また、アクチュエータの特性を図１６に示すように低域ゲインを小さく、位相進み量を大きく、関節の粘性抵抗を小さくするように設定することで、該当する関節部位に機械的受動性（コンプライアンス）と速応性を与えることができる。したがって、（Ｂ）及び（Ｃ）の状態のときに、脚が位置エネルギを半利用するような受動的動作を行なうときに、機械的受動性（コンプライアンス）を持ちつつ高い帯域の追従制御を行なうことが可能となる。また、（Ｄ）の状態のときのように着床の瞬間に機械的受動性（コンプライアンス）を持たせることで、ロボット全体が柔らかい系になるため、機体全体に加わる衝撃を緩和することができる。
【０１９３】
また、図２３には、脚式移動ロボット１００が肩関節ピッチ軸４、体幹ピッチ軸９、股関節ピッチ軸１２、膝関節ピッチ軸１４などの高さ方向に連結された略平行な複数の関節軸からなるリンク構造体としてモデル化して、各関節ピッチ軸を同期強調的に駆動させてうつ伏せ姿勢に向かって転倒していく動作を示している。
【０１９４】
ロボットは、リンク構造体のリンク端である足底のみで立位しているとする（図２３（１））。
【０１９５】
このとき、外力又は外力モーメントの印加により、ＺＭＰ釣合い方程式のモーメント・エラー項Ｔをキャンセルできなくなり、足底のみで形成するＺＭＰ安定領域の外にＺＭＰが逸脱したことに応答して、ＺＭＰを支持多角形に維持したまま転倒動作を開始する。
【０１９６】
転倒動作では、まず、機体の支持多角形の面積Ｓの時間ｔ当たりの変化量ΔＳ／Δｔを最小にするリンクを探索するとともに、手先を含むリンクで変化量ΔＳ／Δｔを最小にするような手先の目標着床点を探索する。そして、選択されたリンクを目標着床点に着床することが機体ハードウェアの制約上（各関節の可動角、各関節アクチュエータのトルク、関節力、角速度、角加速度など）、実行可能かどうかを判別する。
【０１９７】
機体ハードウェア上実行可能である場合には、既に着床している足底リンクに加え、他のリンクが着床する。そして、これら着床リンクで形成される最小支持多角形内にＺＭＰを移動する（図２３（２））。
【０１９８】
次いで、機体ハードウェアが許容する限り、着床点を移動して、支持多角形を拡大していく（図２３（３））。
【０１９９】
そして、各関節の可動角、各関節アクチュエータのトルク、関節力、角速度、角加速度など機体ハードウェアの制約から、もはや着床点を移動することができなくなると、今度は、着床中のリンクに挟まれている離床リンクを着床することができるかどうかを判別する。
【０２００】
機体ハードウェア上、着床リンク間の離床リンクを着床することが可能である場合には、これらを着床して、着床リンク数を増やす（図２３（４））。
【０２０１】
さらに、機体ハードウェアが許容する限り、着床点を移動して、支持多角形を拡大していく（図２３（５））。
【０２０２】
そして最後に、高さ方向に連結された略平行な複数の関節軸からなるリンク構造体の一端側から１以上のリンクと、他端がわから２以上のリンクを離床させて、且つ、それらの中間に位置するリンクを１以上着床させ、さらに足部を着床させた状態で、ＺＭＰを支持多角形内に維持しながら、支持多角形が最大となる姿勢を形成する。この姿勢において、機体の位置エネルギが最小であれば、転倒動作は完了である。
【０２０３】
図２４には、実機が立位姿勢からうつ伏せ姿勢に転倒していく様子を示している。同図に示す例では、機体の後方から外力を受けて後ろ向きに転倒動作を行なう様子を示しているが、Ａ〜Ｄの４段階に分けて動作状況を捉えることができる。
【０２０４】
（Ａ）転倒前の両脚支持の状態
姿勢制御されている状態であり、すべての関節部位のアクチュエータが図１５に示す周波数特性に設定されている。
【０２０５】
この場合、サーボ制御器の比例ゲインを大きくとって、系全体のゲインを大きくし、低周波数帯域までゲインを取れるようにする。また、周波数特性を同図に示すように高域で位相進み量が小さくなるようにし、速応性にはあまり寄与しないが安定性を保証できるようにする。また、モータの粘性抵抗を大きくし、振動などの外乱にもロバストになるようにする。この結果、機体全体として、位置決め精度を優先して振動などの外乱にもロバストとなる。
【０２０６】
【０２０７】
（Ｂ）転倒に至る外力・衝撃を受け、衝撃を最小限にするための転倒動作を開始足首ロール軸、足首ピッチ軸、膝ピッチ軸、肘ピッチ軸など位置エネルギ制御に関わる関節部位のアクチュエータを、図１６に示す周波数特性に設定する。また、それ以外の安定領域制御に関わる関節部位のアクチュエータを、図１５に示す周波数特性に維持する。
【０２０８】
足首ロール軸、足首ピッチ軸、膝ピッチ軸、肘ピッチ軸など位置エネルギ制御に関わる関節部位では、速応性とコンプライアンスが重要となる。そこで、比例ゲインを小さくして、系全体のゲインを小さくすることで、低周波数帯域のゲインを小さくし、機械的受動性（コンプライアンス）を得易くする。また、周波数特性を同図に示すように中高域で位相進み量が大きくなるようにし、速応性を得るようにする。また、モータの粘性抵抗を小さくし、機械的受動性（コンプライアンス）を得易いようにする。
【０２０９】
（Ｃ）足底が離床し、足底面で請けていた床反力がゼロになる
股関節ピッチ軸、肩ピッチ軸など位置エネルギ制御に関わる関節部位のアクチュエータを、さらに図１６に示す周波数特性に設定する。この結果、当該関節部位では機械的受動性（コンプライアンス）と速応性を優先した特性が得られる。また、それ以外の安定領域制御に関わる関節部位のアクチュエータを、図１５に示す周波数特性に維持する。
【０２１０】
（Ｄ）ロボットが背面に着床し、転倒動作が終了
着床と同時に残りすべての関節部位のアクチュエータを図１６に示す周波数特性に設定する。この結果、機体全体として機械的受動性（コンプライアンス）と速応性を優先した特性が得られる。
【０２１１】
アクチュエータの特性を図１５に示すように低域ゲインを大きく、位相進み量を小さく、関節の粘性抵抗を大きくするように設定することで、高精度な位置決めが可能となる。したがって、（Ｂ）及び（Ｃ）の状態のときに、転倒動作を制御する際の規範となるΔＳ／Δｔの量を制御するための主成分となる関節の位置決め精度が確保され、動作の安定性が増す。
【０２１２】
また、アクチュエータの特性を図１６に示すように低域ゲインを小さく、位相進み量を大きく、関節の粘性抵抗を小さくするように設定することで、該当する関節部位に機械的受動性（コンプライアンス）と速応性を与えることができる。したがって、（Ｂ）の状態のときに、脚や腕などが位置エネルギを半利用するような受動的動作を行なうときに、機械的受動性（コンプライアンス）を持ちつつ高い帯域の追従制御を行なうことが可能となる。また、（Ｃ）の状態のときに、脚や腕などが着床後も機体を支えつつ動作を行なう場合に、機械的受動性（コンプライアンス）を利用して機体に加わる衝撃を緩和しつつ高い帯域の追従動作を行なうことが可能となる。また、（Ｄ）の状態のときのように着床の瞬間に機械的受動性（コンプライアンス）を持たせることで、ロボット全体が柔らかい系になるため、機体全体に加わる衝撃を緩和することができる。
【０２１３】
［追補］
以上、特定の実施例を参照しながら、本発明について詳解してきた。しかしながら、本発明の要旨を逸脱しない範囲で当業者が該実施例の修正や代用を成し得ることは自明である。
【０２１４】
本発明の要旨は、必ずしも「ロボット」と称される製品には限定されない。すなわち、電気的若しくは磁気的な作用を用いて人間の動作に似せた運動を行う機械装置であるならば、例えば玩具等のような他の産業分野に属する製品であっても、同様に本発明を適用することができる。
【０２１５】
また、本明細書中では、モータ・コイルへのコイル電流のスイッチング制御にバイポーラ・トランジスタからなるスイッチング素子を用いて構成される回路例について説明したが、ＭＯＳ−ＦＥＴやその他の半導体素子を用いてこの種の制御回路を実装することができることは、当業者には自明である。
【０２１６】
要するに、例示という形態で本発明を開示してきたのであり、本明細書の記載内容を限定的に解釈するべきではない。本発明の要旨を判断するためには、冒頭に記載した特許請求の範囲の欄を参酌すべきである。
【０２１７】
【発明の効果】
以上詳記したように、本発明によれば、転倒・落下の途上における脚部だけでなく胴体及び腕部を含め機体全体の運動制御によりロボットが被る損害を限りなく軽減することができる、優れた脚式移動ロボットを提供することができる。
【０２１８】
また、本発明によれば、転倒・落下の途上における機体全体の運動時においてアクチュエータそのものの特性とアクチュエータの制御器の特性の２つを動的又は静的に制御することにより安定且つ高効率な動作を実現することができる、優れた脚式移動ロボットを提供することができる。
【図面の簡単な説明】
【図１】本発明の実施に供される脚式移動ロボットが直立している様子を前方から眺望した様子を示した図である。
【図２】本発明の実施に供される脚式移動ロボットが直立している様子を後方から眺望した様子を示した図である。
【図３】脚式移動ロボットが具備する関節自由度構成を模式的に示した図である。
【図４】脚式移動ロボット１００の制御システム構成を模式的に示した図である。
【図５】アクチュエータのサーボ制御器の構成を示した図である。
【図６】図５に示したモータと減速機の伝達関数表現モデルゲイン及び位相の周波数特性を示した図である。
【図７】図５に示したサーボ制御器において、位相補償型制御の設計例として位相補償帯域を任意に選択する例を説明するための図である。
【図８】図５に示したサーボ制御器において、位相補償型制御の設計例として位相補償を施す量を任意に選択する例を説明するための図である。
【図９】図５に示したサーボ制御器において、Ｋで示した直列補償ゲインの大小を変更する制御器の設計例を説明するための図である。
【図１０】図７に示したように位相補償量が一定で周波数帯域を任意に選ぶようにアクチュエータのサーボ制御器を実装したときの開ループ特性を示した図である。
【図１１】図１０に示したうち高域でのみ位相進み補償を施した例Ｃ（ｓ）−３においてさらに直列補償ゲインの制御を採り入れた様子を示した図である。
【図１２】コイル電流の制御メカニズムを適用したＤＣモータのコイル電流供給用の電流制御回路の等価回路の構成例を示した図である。
【図１３】付加論理回路の具体的な回路構成を示した図である。
【図１４】ＰＷＭ制御により所定のデューティ比を持つＢＲＡＫＥ＿ＰＷＭ制御信号が入力されたときの付加論理回路の各トランジスタ制御用信号の出力特性を、コイル電流波形特性及びトルク出力特性とともに示した図である。
【図１５】位置決め精度が重要な部分に適用することができるアクチュエータの周波数特性を示した図である。
【図１６】速応性とコンプライアンスが重要となる部分に適用することができるアクチュエータの周波数特性を示した図である。
【図１７】図３に示す多関節自由度構成を具備する脚式移動ロボット１００の多質点近似モデルを示した図である。
【図１８】脚式移動ロボット１００において、ＺＭＰを安定度判別規範に用いて安定歩行可能な機体運動を生成するための処理手順を示したフローチャートである。
【図１９】脚式移動ロボット１００における脚式作業中の機体の動作制御の概略的な処理手順を示したフローチャートである。
【図２０】脚式移動ロボット１００が足部の計画不能により転倒動作を行なうための処理手順を示したフローチャートである。
【図２１】脚式移動ロボット１００を肩関節ピッチ軸４、体幹ピッチ軸９、股関節ピッチ軸１２、膝関節ピッチ軸１４などの高さ方向に連結された略平行な複数の関節軸からなるリンク構造体としてモデル化して、各関節ピッチ軸を同期強調的に駆動させて仰向け姿勢に向かって転倒していく動作を示した図である。
【図２２】実機が立位姿勢から仰向け姿勢に転倒していく様子を示した図である。
【図２３】脚式移動ロボット１００が肩関節ピッチ軸４、体幹ピッチ軸９、股関節ピッチ軸１２、膝関節ピッチ軸１４などの高さ方向に連結された略平行な複数の関節軸からなるリンク構造体としてモデル化して、各関節ピッチ軸を同期強調的に駆動させてうつ伏せ姿勢に向かって転倒していく動作を示した図である。
【図２４】実機が立位姿勢からうつ伏せ姿勢に転倒していく様子を示した図である。
【符号の説明】
１…首関節ヨー軸
２Ａ…第１の首関節ピッチ軸
２Ｂ…第２の首関節（頭）ピッチ軸
３…首関節ロール軸
４…肩関節ピッチ軸
５…肩関節ロール軸
６…上腕ヨー軸
７…肘関節ピッチ軸
８…手首関節ヨー軸
９…体幹ピッチ軸
１０…体幹ロール軸
１１…股関節ヨー軸
１２…股関節ピッチ軸
１３…股関節ロール軸
１４…膝関節ピッチ軸
１５…足首関節ピッチ軸
１６…足首関節ロール軸
３０…頭部ユニット，４０…体幹部ユニット
５０…腕部ユニット，５１…上腕ユニット
５２…肘関節ユニット，５３…前腕ユニット
６０…脚部ユニット，６１…大腿部ユニット
６２…膝関節ユニット，６３…脛部ユニット
８０…制御ユニット，８１…主制御部
８２…周辺回路
９１，９２…接地確認センサ
９３，９４…加速度センサ
９５…姿勢センサ
９６…加速度センサ
１００…脚式移動ロボット[0001]
TECHNICAL FIELD OF THE INVENTION
The present invention relates to a mechanical device of a multi-axis drive system such as a robot, a general-purpose assembly device, a robot / hand device, and other multi-axis control devices, and in particular, a position control comprising a proportional gain of a series compensation and a phase compensation element. The present invention relates to a robot in which each joint part is formed by a servo controller of an actuator constituting a system.
[0002]
More specifically, the present invention relates to a bipedal legged mobile robot in which each axis link is controlled by high gain PD control. In particular, the present invention relates to the characteristics of the actuator itself when the entire body moves during a fall or a fall. The present invention relates to a legged mobile robot that achieves stable and highly efficient operation by dynamically or statically controlling two characteristics of an actuator controller.
[0003]
[Prior art]
A mechanical device that performs a motion resembling a human motion using an electric or magnetic action is called a “robot”. It is said that the root of the robot is derived from the Slavic word "ROBOTA (slave machine)". In Japan, robots began to spread in the late 1960s, but most of them were industrial robots such as manipulators and transfer robots for the purpose of automation and unmanned production work in factories. Met.
[0004]
A stationary robot, such as an arm-type robot, which is implanted and used in a specific place, operates only in a fixed and local work space such as assembling and sorting parts. On the other hand, the mobile robot has a work space that is not limited, and can freely move on a predetermined route or on a non-route to perform a predetermined or arbitrary human work, or perform a human or dog operation. Alternatively, a wide variety of services that can replace other living things can be provided. Among them, legged mobile robots are unstable and difficult to control their posture and walking compared to crawler-type and tire-type robots.However, climbing up and down stairs and ladders, climbing over obstacles, and distinguishing between terrain and rough terrain It is excellent in that a flexible walking / running operation can be realized regardless of the operation.
[0005]
Recently, it has been modeled on pet-type robots that mimic the body mechanisms and movements of quadrupedal animals such as dogs and cats, or the body mechanisms and movements of animals that walk upright on two legs such as humans. Research and development on legged mobile robots, such as designed “humanoid” or “humanoid” robots, has been progressing, and expectations for their practical use are increasing.
[0006]
There have already been proposed a number of techniques relating to posture control and stable walking for a type of robot that performs legged movement by biped walking. Stable “walking” here can be defined as “moving using the legs without falling over”. A bipedal walking robot such as a humanoid has a higher center of gravity and a narrower ZMP stable area during walking than a quadrupedal walking. For this reason, the problem of posture stabilization is particularly important in a bipedal walking robot.
[0007]
The posture stability control of the robot is very important in avoiding the falling of the robot. This is because falling means that the robot interrupts the work being performed, and considerable effort and time is spent in getting up from the falling state and resuming the work. Also, when the robot falls, there is a danger that the robot body itself or the object on the other side that collides with the falling robot may be fatally damaged.
[0008]
During walking, gravity, inertial force, and these moments act on the road surface from the walking system due to gravity and acceleration caused by the walking motion. According to the so-called "Dalambert principle", they balance the floor reaction force and the floor reaction force moment as a reaction from the road surface to the walking system. As a result of the mechanical inference, there is a point where the pitch and roll axis moments are zero, that is, “ZMP (Zero Moment Point)” on or inside the side of the supporting polygon formed by the sole and the road surface. For example, Non-Patent Document 1). Many proposals relating to posture stability control of a legged mobile robot and prevention of falling during walking use this ZMP as a criterion for determining walking stability. The bipedal walking pattern generation based on the ZMP standard has an advantage that a sole landing point can be set in advance, and the kinematic constraint condition of the toe according to the road surface shape can be easily considered. In addition, using ZMP as a stability determination criterion means that a trajectory, not a force, is treated as a target value in motion control, so that technical feasibility is increased.
[0009]
The posture stability control of a robot using ZMP as a stability discrimination standard is basically to search for a point where the moment is zero on or inside the support polygon formed by the sole and the road surface. (For example, see Patent Documents 1 to 5).
[0010]
Of course, in spite of performing the maximum attitude stabilization control so that the aircraft does not fall, the control is inadequate or unexpected external factors (for example, collision with an unexpected object, protrusion or dent on the floor, etc.) Depending on road conditions, appearance of obstacles, etc.), the posture may be lost and the movable leg alone will not be able to support it, and the robot may fall over
.
[0011]
For example, in a situation where a legged mobile robot is likely to fall, damage to the robot due to the fall and damage to an object on the other side with which the robot collides when the fall is possible can be reduced as much as possible. However, the robot merely controls to lower the center of gravity of the robot at the time of landing when falling, and when the vehicle actually falls, in order to minimize damage, not only the legs but also the torso and There is no discussion on how to operate the entire aircraft, including the arms.
[0012]
In addition, a multi-axis drive system mechanical device generally has a large number of joint degrees of freedom, and implements joint movements with an actuator motor. In this case, the rotation position, the rotation amount, and the like of each motor are taken out, and the desired operation pattern can be reproduced and the attitude control can be performed by the servo control. Generally, each joint of each joint viewed from the motion control theory is controlled by the high gain PD control, and the joints generally operate with constant characteristics.
[0013]
However, as can be seen from the results of human motion research, in order to achieve stable and highly efficient motion, it is necessary to locally increase or decrease the force and to increase or decrease the compliance (mechanical passivity) of each joint. is important.
[0014]
When capturing the motion of each axis of the joint as a position control system, it is better to use a high-gain and high-bandwidth servo controller to control so as to reduce the control deviation. In consideration of the action of energy and kinetic energy, it is preferable to simultaneously lower the gain and raise or lower the frequency band for which phase compensation is performed.
[0015]
However, realizing such control on the robot body requires a function of dynamically and statically controlling two characteristics of the actuator itself and the characteristics of the controller of the actuator.
[0016]
For example, there has been proposed a walking control device of a legged mobile robot that can stably walk on a known or unknown walking road surface. In other words, in a bipedal legged mobile robot having a structure similar to a human body with an arm on the upper body, when the frictional force decreases on the walking road surface and the stability decreases, the state is driven and the stability is reduced. Is secured or recovered (for example, see Patent Document 7). However, this is realized by controlling the feed forward gain, and there is no reference to the viscosity or frequency characteristics of the joint and no concept of compliance.
[0017]
[Patent Document 1]
JP-A-5-305579
[Patent Document 2]
JP-A-5-305581
[Patent Document 3]
JP-A-5-305558
[Patent Document 4]
JP-A-5-305585
[Patent Document 5]
JP-A-5-305586
[Patent Document 6]
JP-A-11-48170
[Patent Document 7]
JP-A-7-205069
[Non-patent document 1]
"LEGGED LOCOMOTION ROBOTS" by Miomir Vukobratovic (Ichiro Kato et al., Walking Robots and Artificial Feet, Nikkan Kogyo Shimbun)
[0018]
[Problems to be solved by the invention]
An object of the present invention is to provide an excellent legged mobile robot capable of reducing the damage to the robot as much as possible by controlling the motion of the entire body including the torso and the arms in addition to the legs during the fall and fall. Is to do.
[0019]
A further object of the present invention is to provide a stable and highly efficient operation by dynamically or statically controlling the characteristics of the actuator itself and the characteristics of the controller of the actuator during the movement of the entire body during the fall and fall. An object of the present invention is to provide an excellent legged mobile robot capable of realizing the following.
[0020]
Means and Action for Solving the Problems
The present invention has been made in consideration of the above problems, and is a legged mobile robot including a plurality of joint sites including a plurality of movable legs,
Actuator characteristic control means for performing a combination of gain and phase compensation control of the servo controller of the actuator at each joint site and control of the viscous resistance of the actuator motor,
The actuator characteristic control means includes a first actuator characteristic for increasing the low-frequency gain, a small amount of phase lead, and a high viscous resistance of the joint for each stage of the overturning operation. Switching between a second actuator characteristic that reduces the gain, increases the amount of phase lead, and reduces the viscous resistance of the joint;
It is a legged mobile robot characterized by the following.
[0021]
According to the present invention, in a servo controller of an actuator constituting each joint part of a legged mobile robot, by adjusting a proportional gain and a phase compensation element, positioning accuracy required for each joint part of the robot and mechanical passive Performance (compliance) and operation speed can be arbitrarily set.
[0022]
In addition, when the coil of the actuator / motor is not energized, the coil is intermittently switched to the short-circuit state or the open state, so that the viscous resistance of the motor can be adjusted and the robustness to disturbance such as vibration can be changed.
[0023]
Furthermore, by combining gain and phase compensation control in these actuator servo controllers with control of the viscous resistance of the actuator / motor, the frequency characteristics of the actuator, which can be applied to parts where positioning accuracy is important, The frequency characteristics of the actuator that can be applied to a portion where compliance is important can be obtained.
[0024]
Here, the actuator characteristic control means sets the characteristics of the actuator to “increase the low-frequency gain”, “decrease the amount of phase lead in a high frequency region”, and “increase the viscous resistance of the joint” to thereby increase the actuator characteristic. Accurate positioning control becomes possible, and posture stability increases.
[0025]
Further, the actuator characteristic control means sets the characteristics of the actuator to “small low-frequency gain”, “large phase lead amount”, and “small viscous resistance of the joint” to achieve high mechanical passivity and high speed. Since responsiveness can be provided, it is possible to perform high-bandwidth follow-up control while reducing the impact force at the moment of landing.
[0026]
The legged mobile robot applies force to the aircraft while performing legged work in a standing posture, using a floor reaction force sensor or acceleration sensor located on the sole, or an acceleration sensor located on the waist of the torso. External force is detected. Then, based on these detected external forces, a ZMP balance equation is established, and a ZMP in which the moment applied to the aircraft is balanced on or inside the side of the support polygon formed by the sole and the road surface is set. In addition, the attitude stabilization control of the airframe is performed by always planning the ZMP trajectory.
[0027]
However, the moment error on the ZMP balance equation cannot be canceled due to circumstances such as an excessive external force applied to the fuselage or an unfavorable road surface condition. In some cases, ZMP placement may be difficult or impossible. In such a case, the legged mobile robot according to the present invention gives up the posture stabilization control of the body and performs a predetermined overturning operation to minimize damage to the body at the time of falling to the floor. It has become.
[0028]
That is, when the vehicle falls, a portion where the change ΔS / Δt per unit time Δt of the area S of the support polygon formed by the ground contact point of the vehicle and the road surface is minimized, and the support formed by the ground contact point of the vehicle and the road surface are formed. The target landing point at which the selected part should land is set such that the change ΔS / Δt per unit time Δt of the area S of the polygon is minimized, and the part is landed. Then, the newly formed supporting polygon is further enlarged by landing.
[0029]
Until the potential energy of the fuselage becomes minimum and the overturning operation is completed, an operation of searching for a part where ΔS / Δt is minimum and landing the part at a target landing point where ΔS / Δt is minimum is performed. Or, the operation of enlarging the newly formed support polygon is repeatedly executed.
[0030]
As described above, the change amount ΔS / Δt of the area S of the support polygon per time t is minimized, and the support polygon when the floor falls is maximized, so that the support polygon receives from the floor at the time of fall. The impact can be dispersed throughout the body to minimize damage.
[0031]
Here, the actuator characteristic control means increases the low-frequency gain, the phase lead amount, and the joint viscous resistance by increasing the low-frequency gain, the phase lead amount, and the joint resistance related to the stable region control at each stage of the overturning operation. By maintaining this, it is possible to perform high-precision positioning control of each of these joint parts, and to increase the stability of the posture. As a result, the positioning accuracy of the joint, which is the main component for controlling the amount of ΔS / Δt, which is a reference when controlling the overturning operation of the body, is secured, and the stability of the operation can be increased.
[0032]
Further, the actuator characteristic control means may increase the low-frequency gain, reduce the phase lead amount, and increase the viscous resistance of the joint, for each stage of the overturning operation, by changing the actuator characteristics of each joint site related to the potential energy control. By switching from the first actuator characteristic to the second actuator characteristic that reduces the low-frequency gain, increases the amount of phase lead, and reduces the viscous resistance of the joint, the mechanical passivity (compliance) and the speed of these joint parts are increased. Make it responsive.
[0033]
As a result, when the leg performs a passive operation in which the potential energy is partially used, it is possible to perform a high-bandwidth tracking control while maintaining mechanical passivity (compliance).
[0034]
Furthermore, by switching the characteristics of the actuators at the joints related to shock absorption at the time of landing to the second actuator characteristics that reduce the low-frequency gain, increase the amount of phase lead, and reduce the viscous resistance of the joints, Ensure that the site has mechanical passivity (compliance) and fast response.
[0035]
As a result, by giving mechanical passivity (compliance) at the moment of landing, the entire robot becomes a soft system, so that the impact applied to the entire body can be reduced.
[0036]
Further objects, features, and advantages of the present invention will become apparent from more detailed descriptions based on embodiments of the present invention described below and the accompanying drawings.
[0037]
BEST MODE FOR CARRYING OUT THE INVENTION
Hereinafter, embodiments of the present invention will be described in detail with reference to the drawings.
[0038]
A. Mechanical configuration of a legged mobile robot
FIG. 1 and FIG. 2 show a state in which a “humanoid” or “humanoid” legged mobile robot used for carrying out the present invention stands upright, as viewed from the front and the rear, respectively. As illustrated, the legged mobile robot includes a body, a head, left and right upper limbs, and two left and right lower limbs that perform legged movement. For example, a control unit ( (Not shown) controls the operation of the aircraft in an integrated manner.
[0039]
Each of the left and right lower limbs includes a thigh, a knee joint, a shin, an ankle, and a foot, and is connected at a substantially lowermost end of the trunk by a hip joint. Each of the left and right upper limbs is composed of an upper arm, an elbow joint, and a forearm, and is connected at the left and right side edges above the trunk by a shoulder joint. The head is connected to the center of the uppermost end of the trunk by a neck joint.
[0040]
The control unit is equipped with a controller (main control unit) that processes external controls from various sensors (described later) and drive control of each joint actuator that constitutes this legged mobile robot, and a power supply circuit and other peripheral devices. It is a housing. The control unit may include a communication interface and a communication device for remote control.
[0041]
The legged mobile robot thus configured can realize bipedal walking by controlling the whole body in a coordinated manner by the control unit. In general, such bipedal walking is performed by repeating a walking cycle divided into the following operation periods. That is,
(1) Right leg lifted, left leg supporting single leg
(2) Both legs supported period with right foot touching down
(3) Single leg support period with right leg with left leg lifted
(4) Both legs supported period when left foot touched down
[0042]
Walking control in the legged mobile robot is realized by planning a target trajectory of the lower limb in advance and correcting the planned trajectory in each of the above periods. That is, in the two-leg support period, the correction of the lower limb trajectory is stopped, and the waist height is corrected to a constant value using the total correction amount for the planned trajectory. In the single leg support period, a corrected trajectory is generated so that the relative positional relationship between the ankle and the waist of the corrected leg is returned to the planned trajectory.
[0043]
In general, the attitude stabilization control of the body, including the trajectory correction of the walking motion, is performed by interpolation calculation using a fifth-order polynomial so that the position, velocity, and acceleration for reducing the deviation from ZMP are continuous. . ZMP (Zero Moment Point) is used as a criterion for determining the stability of walking. The stability discrimination standard based on ZMP is based on the principle of “Dallambert” that gravity and inertia force from the walking system to the road surface and these moments balance the floor reaction force and the floor reaction force moment as a reaction from the road surface to the walking system. based on. As a result of the mechanical inference, the point where the pitch axis and the roll axis moment are zero on or inside the side of the support polygon (that is, the ZMP stable area) formed by the sole and the road surface, that is, "ZMP (Zero Moment)" Point) "exists.
[0044]
FIG. 3 schematically shows a configuration of a degree of freedom of a joint included in the legged mobile robot. As shown in the figure, the legged mobile robot has an upper limb including two arms and a head 1, a lower limb including two legs for realizing a moving operation, and a body connecting the upper limb and the lower limb. This is a structure having a plurality of limbs, which is composed of a trunk.
[0045]
The neck joint (Neck) that supports the head has four degrees of freedom: a neck joint yaw axis 1, first and second neck joint pitch axes 2A and 2B, and a neck joint roll axis 3.
[0046]
Each arm has, as its degrees of freedom, a shoulder joint pitch axis 4 at the shoulder (Shoulder), a shoulder joint roll axis 5, an upper arm yaw axis 6, an elbow joint pitch axis 7 at the elbow (Elbow), and a wrist ( (Wrist) and a hand part. The hand is actually a multi-joint / multi-degree-of-freedom structure including a plurality of fingers.
[0047]
The trunk (Trunk) has two degrees of freedom: a trunk pitch axis 9 and a trunk roll axis 10.
[0048]
Further, each leg constituting the lower limb includes a hip joint yaw axis 11, a hip joint pitch axis 12, a hip joint roll axis 13, a knee joint pitch axis 14 in a knee (Knee), and an ankle (Ankle) in a hip joint (Hip). ), An ankle joint pitch axis 15, an ankle joint roll axis 16, and a foot.
[0049]
B. Control system configuration of legged mobile robot
FIG. 4 schematically shows a control system configuration of the legged mobile robot 100. As shown in the figure, the legged mobile robot 100 has mechanism units 30, 40, 50R / L and 60R / L representing human limbs, and adaptive control for realizing a cooperative operation between the mechanism units. (However, each of R and L is a suffix indicating each of right and left. The same applies hereinafter).
[0050]
The operation of the entire legged mobile robot 100 is totally controlled by the control unit 80. The control unit 80 transmits and receives data and commands to and from a main control unit 81 composed of main circuit components (not shown) such as a CPU (Central Processing Unit) and a memory, and a power supply circuit and each component of the robot 100. A peripheral circuit 82 including an interface (both not shown) and the like are provided.
[0051]
In realizing the present invention, the installation location of the control unit 80 is not particularly limited. In FIG. 4, it is mounted on the trunk unit 40, but may be mounted on the head unit 30. Alternatively, the control unit 80 may be provided outside the legged mobile robot 100 to communicate with the body of the legged mobile robot 100 by wire or wirelessly.
[0052]
Each joint degree of freedom in the legged mobile robot 100 shown in FIG. 3 is realized by an actuator corresponding to each joint. That is, the head unit 30 includes a neck joint yaw axis actuator A representing each of the neck joint yaw axis 1, the first and second neck joint pitch axes 2A and 2B, and the neck joint roll axis 3.₁, First and second neck joint pitch axis actuators A_2A, A_2B, Neck joint roll axis actuator A_ThreeAre arranged respectively.
[0053]
The trunk unit 40 includes a trunk pitch axis actuator A representing each of the trunk pitch axis 9 and the trunk roll axis 10.₉, Trunk roll axis actuator A_TenIs deployed.
[0054]
The arm unit 50R / L is subdivided into an upper arm unit 51R / L, an elbow joint unit 52R / L, and a forearm unit 53R / L. The shoulder joint pitch axis 4, shoulder joint roll axis 5, upper arm Shoulder joint pitch axis actuator A expressing each of yaw axis 6, elbow joint pitch axis 7, and wrist joint yaw axis 8_Four, Shoulder joint roll axis actuator A_Five, Upper arm yaw axis actuator A₆, Elbow joint pitch axis actuator A₇, Wrist joint yaw axis actuator A₈Is deployed.
[0055]
The leg unit 60R / L is subdivided into a thigh unit 61R / L, a knee unit 62R / L, and a shin unit 63R / L, and the hip joint yaw axis 11, the hip joint pitch axis 12, and the hip joint Hip joint yaw axis actuator A expressing each of roll axis 13, knee joint pitch axis 14, ankle joint pitch axis 15, and ankle joint roll axis 16₁₁, Hip joint pitch axis actuator A₁₂, Hip roll axis actuator A₁₃, Knee joint pitch axis actuator A₁₄, Ankle joint pitch axis actuator A_Fifteen, Ankle joint roll axis actuator A₁₆Is deployed.
[0056]
Actuator A used for each joint₁, A_Two, A_Three... can be constituted by a small-sized AC servo actuator (described above) of a type directly connected to a gear and a type in which a servo control system is integrated into one chip and mounted in a motor unit.
[0057]
Sub-control units 35, 45, 55, and 65 for actuator drive control are provided for each mechanism unit such as the head unit 30, the trunk unit 40, the arm unit 50, and the leg units 60.
[0058]
An acceleration sensor 95 and a posture sensor 96 are provided on the trunk 40 of the body. The acceleration sensor 95 is disposed in each of the X, Y, and Z axis directions. By arranging the acceleration sensor 95 on the waist of the body, the waist, which is a part where the mass operation amount is large, is set as the control target point, and the posture and acceleration at that position are directly measured, and the posture stability control based on ZMP is performed. Can be performed.
[0059]
In addition, grounding confirmation sensors 91 and 92 and acceleration sensors 93 and 94 are provided on the legs 60R and 60L, respectively. The ground contact confirmation sensors 91 and 92 are configured by attaching a pressure sensor to the sole, for example, and can detect whether the sole has landed on the basis of the presence or absence of a floor reaction force. The acceleration sensors 93 and 94 are arranged at least in the X and Y axis directions. By arranging the acceleration sensors 93 and 94 on the left and right feet, the ZMP equation can be directly assembled with the feet closest to the ZMP position.
[0060]
The main control unit 80 can dynamically correct the control target in response to the outputs of the sensors 91 to 96. More specifically, a whole body motion pattern in which the upper limb, the trunk, and the lower limb of the legged mobile robot 100 are driven cooperatively by performing adaptive control on each of the sub-control units 35, 45, 55, 65. To achieve.
[0061]
The whole body movement of the robot 100 on the body sets foot movement, ZMP (Zero Moment Point) trajectory, trunk movement, upper limb movement, waist height, and the like, and instructs an operation according to these set contents. The command is transferred to each of the sub-control units 35, 45, 55, 65. Each of the sub-control units 35, 45,... Interprets the command received from the main control unit 81 and₁, A_Two, A_Three, And outputs a drive control signal. The “ZMP” referred to here is a point on the floor at which the moment due to the floor reaction force during walking becomes zero, and the “ZMP trajectory” is, for example, a ZMP trajectory during the walking operation of the robot 100. Means a moving trajectory.
[0062]
C. Control of actuator characteristics
C-1. Actuator operating speed, mechanical passivity
Japanese Patent Application No. 2001-233691 already assigned to the present applicant discloses a legged mobile robot that performs a high-speed switching operation between a closed link state and an open link state with respect to the outside world or a work object. I have. That is, in a robot having limbs composed of one or more rotary joints (each joint may have two or more degrees of freedom), the minimum required degree of passive freedom for eliminating dynamic closing errors in each limb. (Such as backlash of the reducer), and properly manage the movable range of each limb. Even if the actuator that drives the joint does not have means for acquiring torque information, high-speed switching operation between the closed link state and the open link state is stably realized.
[0063]
This means that in a bipedal walking robot, a geared motor with a small amount of backlash is arranged at a position near the waist reference coordinates, and a geared motor with a large amount of backlash is arranged at a position near the hands and feet to obtain optimal characteristics. Things. As a second method for optimizing the robot, the position error deviation is controlled by arbitrarily adjusting the open loop gain of the position servo compensator in each joint axis actuator. That is, when the backlash amount is uniform, the servo deviation based on the magnitude of the servo gain is regarded as the backlash amount and is controlled.
[0064]
In the present embodiment, such an optimization method for a legged mobile robot is further extended to adjust not only the proportional gain of the servo controller but also the phase compensation element at each joint.
[0065]
FIG. 5 shows the configuration of the servo controller of the actuator. As shown in the figure, the servo controller has two control elements, a proportional gain K of series compensation and a phase compensation element C (s), and adjusts not only the proportional gain but also the phase compensation element at each joint. FIG. 6 shows frequency characteristics of transfer function expression model gain and phase of the motor and the speed reducer shown in FIG.
[0066]
Here, the phase compensation element is represented by the following equation. Here, n and m are arbitrary natural numbers, and a_i, B_iIs an arbitrary real number, and corresponds to a feedback gain when expressing a state variable. S is a Laplace operator.
[0067]
(Equation 1)

[0068]
The transfer function expression model G (s) of the motor and the speed reducer is represented by the following equation. Where K is the motor gain, J is the moment of inertia of the motor, and D is the viscous drag coefficient of the motor.
[0069]
(Equation 2)

[0070]
First, in the servo controller shown in FIG. 5, an example of arbitrarily selecting a phase compensation band (arbitrarily selecting a frequency band with a constant phase compensation amount) as a design example of phase compensation control will be described with reference to FIG. I will explain it. In the figure,
[0071]
{Circle around (1)} C (s) -1: A gain of about +5.6 dB and a phase lead of about +18 deg are given in a band of 1.0 to 100 Hz.
{Circle around (2)} C (s) -2: Provides about +5.6 dB of gain amplification and about +18 deg of phase advance in a band of 0.1 to 10 Hz.
{Circle around (3)} C (s) -3: A gain of about +5.6 dB and a phase lead of about +18 deg are given in a band of 10 to 1 kHz.
[0072]
As described above, by arbitrarily selecting the frequency band to be subjected to the phase compensation, it is possible to freely set the frequency characteristics of the actuator. Therefore, the joint axis of the robot constituted by such an actuator can dynamically adjust the frequency characteristic in accordance with the attitude of the body and the aspect of operation.
[0073]
Although the example of phase lead compensation is shown in the example shown in FIG. 7, an arbitrary amount of phase delay can be set in an arbitrary frequency band similarly in the case of phase delay compensation.
[0074]
Next, FIG. 8 shows an example of arbitrarily selecting an amount of phase compensation as a design example of phase compensation control in the servo controller shown in FIG. 5 (arbitrarily selecting an amount of phase compensation with a constant frequency band). It will be described with reference to FIG. In the figure,
[0075]
{Circle around (4)} C (s) -4: A gain of about +3.5 dB and a phase lead of about +12 deg are provided in a band of 4.0 to 70 Hz.
(5) C (s) -5: A gain of about +5.6 dB and a phase lead of about +18 deg are provided in a band of 2.0 to 70 Hz.
(6) C (s) -6: A gain of about +6.5 dB and a phase lead of about +21 deg are provided in a band of 1.0 to 70 Hz.
[0076]
Thus, by arbitrarily selecting the amount of phase compensation, the frequency characteristics of the actuator can be set freely. Therefore, the joint axis of the robot constituted by such an actuator can dynamically adjust the frequency characteristic in accordance with the attitude of the body and the aspect of operation.
[0077]
Although the example of phase lead compensation is shown in the example shown in FIG. 8, an arbitrary amount of phase delay can be set in an arbitrary frequency band similarly in the case of phase delay compensation.
[0078]
Next, with reference to FIG. 9, a description will be given of a design example of a controller for changing the magnitude of the series compensation gain indicated by K in the servo controller shown in FIG. This figure corresponds to a case where K is raised and lowered by ± 3 dB in FIG. As shown, the magnitude of the series compensation gain can be arbitrarily set.
[0079]
In order to apply the contents shown in FIGS. 7 to 9 to the actuator for driving the joint axis of the robot, a communication protocol for dynamically or statically changing the parameters constituting these controllers is implemented. Thereby, various characteristics can be given to each joint axis of the robot.
[0080]
Next, the characteristics of the actuator when the servo controller of the actuator having these characteristics is mounted will be described.
[0081]
FIG. 10 shows an open-loop characteristic when the servo controller of the actuator is mounted such that the phase compensation amount is constant and the frequency band is arbitrarily selected as shown in FIG.
[0082]
{Circle around (1)} C (s) -1: Gives about +5.6 dB of gain amplification and about +18 deg of phase lead in a band of 1.0 to 100 Hz.
→ Since the gain is increased as a whole, positioning accuracy and followability are improved, but energy loss tends to occur. In addition, the load may become unstable when the load increases.
{Circle around (2)} C (s) -2: Provides about +5.6 dB gain amplification and about +18 deg phase lead in the 0.1 to 10 Hz band.
→ Has characteristics intermediate between C (s) -1 and C (s) -2.
{Circle around (3)} C (s) -3: Provides about +5.6 dB of gain amplification and about +18 deg of phase lead in the band of 10 to 1 kHz.
→ Since the phase lead is compensated only in the high frequency range, the effect is not so much seen when moving slowly, but it is effective when moving fast, such as running, flying, and dancing.
[0083]
As described above, by arbitrarily selecting the frequency band to be subjected to the phase compensation, it is possible to freely set the frequency characteristics of the actuator. Therefore, the joint axis of the robot constituted by such an actuator can dynamically adjust the frequency characteristic in accordance with the attitude of the body and the aspect of operation.
[0084]
FIG. 11 shows a state in which the control of the series compensation gain is further adopted in the example C (s) -3 in which the phase lead compensation is performed only in the high frequency band shown in FIG. In this case, as in the example shown in FIG. 5, the gain increases and decreases in the same phase.
[0085]
In the example shown in FIG. 10, the phase compensation example C (s) -3 has little effect at the time of slow operation, but as shown in FIG. 11, by increasing the gain in the low frequency band, Control deviation can be reduced. As a result, it is possible to respond to the command value with a small delay even during a slow operation.
[0086]
The mechanism for adjusting not only the proportional gain of the servo controller but also the phase compensation element at each joint part in the actuator servo controller has been described above. This makes it possible to locally increase or decrease the force and to increase or decrease the compliance (mechanical passivity) of each joint in order to realize a stable and highly efficient operation.
[0087]
For example, when the motion of each axis of the joint is captured as a position control system, it is better to use a high-gain and high-bandwidth servo controller to control so that the control deviation is reduced. In consideration of the effects of potential energy and kinetic energy, it is preferable to simultaneously lower the gain and raise or lower the frequency band for which phase compensation is performed.
[0088]
C-2. Viscous resistance of actuator / motor
In addition to the characteristics such as the operation speed and mechanical passivity of the actuator as described in the above section C-1, a method of variably controlling the viscous resistance of the actuator can be adopted.
[0089]
For example, a motor of a type that generates a rotational torque by controlling a supply current to a coil to form a predetermined magnetic flux distribution generally includes a first transistor switch group that connects a coil terminal to a power supply voltage, and a coil. By driving a switching operation circuit including a second transistor / switch group that grounds the terminal by PWM control, the coil current is controlled to obtain a desired torque or a rotational position, a rotational speed, and the like.
[0090]
Here, at the timing when the motor coil is in the open state during the period when the motor coil is not energized, the current (strictly speaking, electric charge) applied to the motor coil is lost, resulting in loss of torque. In addition, it is easily affected by uneven torque due to cogging.
[0091]
In such a case, even when the motor coil is not energized, a short-circuit state in which the coil does not open is formed, so that the current (strictly, electric charge) applied to the motor coil does not escape. You can do so. At this time, a back electromotive force is generated in the coil of the motor due to the magnetic flux density from the permanent magnet side. A force acts in a direction opposite to the rotation direction of the motor by the back electromotive force, so that a viscous resistance against rotation by an external force can be created, and an effect similar to a brake can be obtained. Due to such viscous resistance to the motor, there is no torque loss and the effect of uneven torque due to cogging is reduced.
[0092]
On the other hand, when the motor is not energized, when such a short-circuit state of the coil is formed, as described above, a kind of viscous resistance can be given to the motor, but when such a motor is used in a robot, Due to the effect of the brake due to the coil short, there is a problem that the compliance (mechanical passivity) is lost.
[0093]
Therefore, by adjusting the ratio between the open state and the short-circuit state of the coil when the motor coil is not energized in accordance with the desired mechanical characteristics, the motor coil at the timing when the motor coil enters the open state is adjusted. Compliance (mechanical passive) due to the problem of torque loss due to the loss of the supplied current (strictly speaking, electric charge) and torque unevenness due to cogging, and the effect of braking due to coil shorts when the motor coil is not energized Gender) can be solved together.
[0094]
Here, the ratio between the energized state and the non-energized state of the motor / coil can be realized by PWM control, and the ratio between the open state and the short-circuit state of the coil in the non-energized state of the motor / coil is also controlled by the PWM control. It can be realized using.
[0095]
FIG. 12 shows a configuration example of an equivalent circuit of a current control circuit for supplying a coil current of a DC motor to which a coil current control mechanism is applied.
[0096]
The current control circuit shown in the figure has a full-bridge configuration, in which a pnp-type transistor A 'and an npn-type transistor A are connected in a forward direction, and a pnp-type transistor B' and an npn-type transistor B are also connected in a forward direction. The circuit connected in the direction_ccAnd the ground GND, and an intermediate point between the transistors A ′ and A and an intermediate point between the transistors B ′ and B are connected by a single-phase coil of the stator.
[0097]
By turning on transistors A 'and B and turning off transistors A and B', a current I_mFlows. By turning off the transistors A ′ and B, the coil is opened and the current I_mWill not flow. By turning off the transistors A 'and B and turning on the transistors A' and B ', the motor coil is short-circuited.
[0098]
The PWM control logic circuit generates current commands to the coil based on a current axis current command (or torque command) from a central control unit (not shown), and performs switching control of each transistor in a PWM system based on these current commands. . That is, the transistors A ′ and B ′ are turned on and the transistors A and B ′ are turned off, and the coil current I_m, And a non-energization period in which the transistors A ′ and B are turned off and the coil is de-energized.
[0099]
The additional logic circuit further includes a signal A output from the PWM control logic circuit.₀And A₀', Row B₀And B₀The control logic for controlling the ON / OFF operation is switched by additional logic. Thus, the switching operation between the open state and the short-circuit state of the coil when the motor coil is not energized is performed. FIG. 13 shows a specific circuit configuration of the additional logic circuit.
[0100]
Signal A for controlling transistor A 'from the PWM control logic circuit₀'And transistor B₀'Control signal B₀And the signal A for controlling the transistor A₀And the signal B for controlling the transistor B₀And the inverted logical product of these logical operation values is ORed with the BRAKE_PWM control signal. The result of the logical sum of the result of this logical sum and the original transistor control signal is the final transistor control signal.
[0101]
When a high level is input to the BRAKE_PWM control signal, the additional logic circuit switches the transistor control signal so as to short-circuit the coil when the coil is not energized. When the normal coil is not energized, the PWM control logic circuit outputs a control signal A₀'And A₀Is high, B₀'And B₀Is set to low, a transistor control signal is output. On the other hand, when the high-level BRAKE_PWM control signal is input, the additional logic circuit₁'And A₁To a low to create a short circuit condition for the coil.
[0102]
On the other hand, when the BRAKE_PWM control signal is in the low state, the additional logic circuit outputs the transistor control signal from the PWM control logic circuit as it is when the coil is not energized, so that the coil is in the open state when the current is not energized.
[0103]
FIG. 14 shows output characteristics of each transistor control signal of the additional logic circuit when a BRAKE_PWM control signal having a predetermined duty ratio is input by PWM control, together with a coil current waveform characteristic and a torque output characteristic.
[0104]
If the coil is short-circuited when the coil is not energized, the time required for the coil current to return to zero becomes longer due to the transient response, but if the coil is opened, the time becomes shorter. The transient response characteristic when the coil is not energized is a mixture of these characteristics according to the duty ratio of the BRAKE_PWM control signal.
[0105]
Therefore, as shown in the figure, when the switching operation of coil energization and coil short circuit is repeated, the next energization is started before the coil current returns to zero when the coil is not energized. The maximum current of the coil at this time gradually increases each time the coil is energized and de-energized, and the increasing tendency is substantially proportional to the duty ratio, that is, the ratio at which the BRAKE_PWM control signal becomes a high level. Similarly, the effective value of the coil current gradually increases as shown in the figure, but its increasing tendency is almost proportional to the duty ratio, that is, the ratio of the BRAKE_PWM control signal at a high level.
[0106]
Further, the output torque T of the motor is obtained by adding the torque constant K of the motor to the coil current._t(T = K)_t(I) As can be seen from the figure, when coil energization and de-energization are repeated, the effective value of the motor torque increases as the coil current increases. The rising tendency at this time is substantially proportional to the duty ratio of the BRAKE_PWM control signal, that is, the ratio at which the control signal goes high. The characteristic that the output of the motor torque increases corresponds to the viscosity coefficient of the motor. In other words, it is possible to dynamically control the viscous resistance of the motor by the duty ratio of the BRAKE_PWM control signal.
[0107]
In this way, by performing PWM control on the duty ratio of the BRAKE_PWM control signal supplied to the additional logic circuit by the PWM control logic circuit, the ratio of the open state and the short-circuit state of the coil when the motor coil is not energized can be set to a desired value. It can be adjusted according to the mechanical properties.
[0108]
Therefore, there is a problem of torque loss due to a loss of current (strictly, electric charge) supplied to the motor coil at a timing when the motor coil is in an open state, uneven torque due to cogging, and non-energization of the motor coil. The problem that the compliance (mechanical passivity) is lost due to the influence of the brake due to the coil short circuit at the time can be solved together.
[0109]
In the above description, a DC motor has been described as an example. However, the same applies to a three-phase motor or a motor of a type that generates a rotational torque by controlling a supply current to another coil to form a predetermined magnetic flux distribution. In addition, the desired viscous resistance of the motor can be obtained by intermittently switching the motor coil when the power is not supplied to the open state and the short-circuit state.
[0110]
C-3. Application to a legged mobile robot
Next, a bipedal legged mobile robot in which the characteristic control mechanism of the actuator servo controller and the characteristic control mechanism of the actuator itself according to the present embodiment are applied to each joint site will be described.
[0111]
As described above, in the servo controller of the actuator, by adjusting the proportional gain and the phase compensation element, the positioning accuracy, mechanical passivity (compliance), and operation speed required at each joint of the robot are arbitrarily set. be able to. In addition, when the coil of the actuator / motor is not energized, the coil is intermittently switched to the short-circuit state or the open state, so that the viscous resistance of the motor can be adjusted and the robustness to disturbance such as vibration can be changed.
[0112]
Furthermore, by combining gain and phase compensation control in these actuator servo controllers with control of the viscous resistance of the actuator / motor, the frequency characteristics of the actuator, which can be applied to parts where positioning accuracy is important, The frequency characteristics of the actuator that can be applied to a portion where compliance is important can be obtained.
[0113]
FIG. 15 shows frequency characteristics of an actuator that can be applied to a portion where positioning accuracy is important.
[0114]
In this case, the proportional gain of the servo controller is increased to increase the gain of the entire system so that the gain can be obtained up to a low frequency band. Further, as shown in the figure, the frequency characteristics are set so that the amount of phase lead is small in a high frequency range, and the stability can be guaranteed while not contributing much to quick response. Further, the viscous resistance of the motor is increased so as to be robust against disturbance such as vibration. In short, the characteristics shown in the figure are characteristics that are robust to disturbances such as vibrations with priority given to positioning accuracy.
[0115]
FIG. 16 shows frequency characteristics of an actuator that can be applied to a portion where quick response and compliance are important.
[0116]
In this case, by reducing the proportional gain to reduce the gain of the entire system, the gain in the low frequency band is reduced, and mechanical passivity (compliance) is easily obtained. Also, as shown in the figure, the amount of phase lead is increased in the middle and high frequency ranges so as to obtain quick response. Further, the viscous resistance of the motor is reduced so that mechanical passivity (compliance) can be easily obtained. In short, the characteristics shown in the figure are characteristics that prioritize mechanical passivity (compliance) and quick response.
[0117]
The degrees of freedom of the joints of the legged mobile robot shown in FIGS. 1 to 3 are realized by the actuator including the servo controller described above. A basic control example of the gain / phase compensation characteristics of the actuator used in each joint part will be described in detail below.
[0118]
(1) Characteristics of actuator applied to neck
At the neck, the proportional gain is set high to give priority to positioning accuracy. In addition, the amount of phase advance is set to a small value so as to maintain the operating speed and not to impair the stability of the increased proportional gain. Further, in order to obtain robustness against vibration disturbance generated during operation of a portion below the torso, the viscous resistance of the joint is set large.
[0119]
(2) Characteristics of actuator applied to shoulder / elbow
When performing a continuous operation such as walking or dancing, the actuator is given such characteristics that mechanical passivity is higher than positioning characteristics. To make the movement more passive, the viscous resistance of the joint is reduced. Further, the proportional gain is set low in order to give the operation a passivity and reduce the energy consumption. Further, in order to increase the operation speed, the frequency band in which phase lead compensation is performed is set high, and the amount of phase lead is set large. Depending on the operation, there are times when only reciprocating motion is performed like a pendulum. In that case, the viscous resistance of the joint and the proportional gain are minimized to obtain mechanical passivity (compliance), and mechanical energy is easily used for operation.
[0120]
On the other hand, when an operation using a force such as pushing or pulling an object is performed, control is performed such that the characteristic of prioritizing positioning accuracy and the characteristic of mechanical passivity are dynamically exchanged according to the load torque value. To generate more force with respect to the load torque value, the proportional gain is increased and the viscous resistance of the joint is increased. In addition, when performing an operation of imitating a constant load with respect to the load torque value, in addition to the adjustment based on the position command value from the host, the proportional gain according to the load torque detected by the actuator internal torque sensor is used. To reduce the viscous resistance of the joint so as to obtain mechanical passivity (compliance).
[0121]
(3) Characteristics of actuator applied to trunk
In order to obtain robustness against vibration disturbance due to own movement, the viscous resistance of the joint is increased. Alternatively, in order to give priority to positioning accuracy, the proportional gain is set high. Alternatively, the amount of phase advance is set small so as not to impair the stability of increasing the proportional gain while maintaining the operation speed.
[0122]
(4) Characteristics of actuator applied to hip joint
In order to obtain robustness against vibration disturbance due to own movement, the viscous resistance of the joint is increased. Alternatively, in order to give priority to positioning accuracy, the proportional gain is set high. Alternatively, the amount of phase advance is set small so as not to impair the stability of increasing the proportional gain while maintaining the operation speed.
[0123]
(5) Characteristics of actuator applied to knee
At the time of a free leg and at the moment of landing, control is performed so that mechanical passivity is higher than positioning accuracy. To make the movement more passive, the viscous resistance of the joint is reduced. Also, the proportional gain is set low in order to make the operation passive and reduce the energy consumption. Further, in order to increase the operation speed, the frequency region in which phase lead compensation is performed is set high, and the amount of phase lead is set large.
[0124]
On the other hand, at the time of the support leg, the characteristic is controlled so that the positioning accuracy is higher than the mechanical passivity. In order to obtain robustness against vibration disturbance due to own movement, the viscous resistance of the joint is increased. Alternatively, in order to give priority to positioning accuracy, the proportional gain is set high. Alternatively, the amount of phase advance is set small so as not to impair the stability of increasing the proportional gain while maintaining the operation speed.
[0125]
(6) Characteristics of actuator applied to ankle
At the time of a free leg and at the moment of landing, control is performed so that mechanical passivity is higher than positioning accuracy. In order to reduce the impact due to the landing on the ankle, the joint viscosity is set small to obtain mechanical passivity (compliance). Further, in order to reduce the impact due to the landing on the ankle, the proportional gain is set low to obtain mechanical passivity (compliance). Further, in order to increase the operation speed, the frequency region in which phase compensation is performed is set high, and the amount of phase advance is set large.
[0126]
On the other hand, at the time of the supporting leg, the viscous resistance of the joint is increased in order to increase the torque generated at the ankle portion and to obtain robustness against vibration disturbance due to the operation of the user. Further, in order to improve the positioning accuracy of the ankle portion, the proportional gain is set high. Further, while maintaining the operation speed, the amount of phase advance is set to be small so as not to impair the stability of the increased proportional gain.
[0127]
D. Posture stability control of a legged mobile robot
Next, in the legged mobile robot 100 according to the present embodiment, the posture stabilization processing at the time of legged work, that is, the posture stabilization processing at the time of performing the whole body cooperative movement including the foot, waist, trunk, and lower limb movements is performed. The procedure will be described.
[0128]
For the posture stability control, ZMP is used for the posture stability control. In the present embodiment, a region where the mass operation amount is maximum, for example, a waist, is set as a local coordinate origin as a control target point on the robot body. Then, a measuring means such as an acceleration sensor is arranged at the control target point, and the posture and acceleration at that position are directly measured to perform posture stability control based on ZMP. Further, by arranging an acceleration sensor on the foot which is a contact portion with the road surface, the local coordinates used for control and the coordinates are directly measured, and the ZMP equation is directly assembled with the foot closest to the ZMP position.
[0129]
D-1. Introduction of ZMP equation
The legged mobile robot 100 according to the present embodiment is an aggregate of infinite, that is, continuous mass points. Here, the amount of calculation for the stabilization process is reduced by replacing the model with an approximate model having a finite number of discrete mass points. More specifically, the legged mobile robot 100 physically having the multi-joint degree of freedom configuration shown in FIG. 3 is handled by replacing it with a multi-mass point approximation model as shown in FIG. The illustrated approximation model is a linear and non-interfering multi-mass approximation model.
[0130]
10, the O-XYZ coordinate system represents roll, pitch, and yaw axes in an absolute coordinate system, and the O'-X'Y'Z 'coordinate system represents roll, pitch, and the like in a motion coordinate system that moves with the robot 100. Each yaw axis is shown. However, the meanings of the parameters in the figure are as follows. Also, symbols with a dash (') should be understood to describe a motion coordinate system.
[0131]
(Equation 3)

[0132]
In the multi-mass model shown in FIG. 17, i is a subscript representing the i-th given mass, and m_iIs the mass of the ith mass point, r_i'Represents the position vector (the motion coordinate system) of the i-th mass point. The machine center of gravity of the legged mobile robot 100 according to the present embodiment exists near the waist. In other words, the waist is the mass point at which the mass operation amount becomes the maximum. In FIG. 17, the mass is mh, and the position vector (however, the motion coordinate system) is r ′._h(R '_hx, R '_hy, R '_hz). In addition, the position vector of the ZMP of the aircraft (the motion coordinate system) is represented by r ′_zmp(R '_zmpx, R '_zmpy, R '_zmpz). In addition, each rotation angle (θ_hx, Θ_hy, Θ_hz) Defines the posture of the waist of the legged mobile robot 100, that is, the rotation of the roll, pitch, and yaw axes.
[0133]
The world coordinate system O-XYZ is an absolute coordinate system and is invariable. In the legged mobile robot 100 according to the present embodiment, acceleration sensors 93, 94, and 96 are disposed on the lower back and the legs of both legs, respectively._qIs detected directly. On the other hand, in the motion coordinate system, that is, in the local coordinate system of the body, O-X'Y'Z 'moves together with the robot.
[0134]
The ZMP equation of the fuselage describes the balance of each moment applied at the control target point. As shown in FIG._iWhen these are set as control target points, all control target points m_iIs a ZMP equation.
[0135]
The ZMP equation of the aircraft described in the world coordinate system (O-XYZ) and the local coordinate system (O-X'Y'Z ') of the aircraft are as follows.
[0136]
(Equation 4)

[0137]
The above equation is for each mass m_iAround the ZMP generated by the acceleration component applied at (radius r_i-R_zmp) And the external force moment M applied to each mass point mi._iAnd the external force F_k(The k-th external force F)_kThe action point of s_k) Are balanced.
[0138]
This ZMP balance equation includes a total moment compensation amount, that is, a moment error component T. By keeping this moment error at zero or within a predetermined tolerance, the attitude stability of the aircraft is maintained. In other words, correcting the body motion (foot motion and the trajectory of each part of the upper body) so that the moment error becomes zero or less than the allowable value is the essence of the posture stability control using ZMP as a stability discrimination criterion. is there.
[0139]
In the present embodiment, since the acceleration sensors 96, 93 and 94 are provided on the waist and the left and right feet, respectively, the ZMP is directly and accurately performed using the acceleration measurement results at these control target points. A balance equation can be derived. As a result, high-speed and more strict posture stability control can be realized.
[0140]
D-2. Whole-body coordinated posture stabilization control
FIG. 18 shows, in the form of a flowchart, a processing procedure in the legged mobile robot 100 for generating a body motion capable of stable walking using the ZMP as a stability determination criterion. However, in the following description, each joint position and motion of the legged mobile robot 100 is described using a linear / non-interfering multi-mass point approximation model as shown in FIG.
[0141]
First, the setting of the foot motion is performed (step S1). The foot motion is motion data in which two or more aircraft poses are connected in chronological order. The motion data is composed of, for example, joint space information indicating displacement of each joint angle of the foot and Cartesian space information indicating joint positions.
[0142]
Next, a ZMP stable area is calculated based on the set foot motion (step S2). The ZMP is a point at which the moment applied to the airframe becomes zero, and basically exists on or inside the side of the supporting polygon formed by the sole contact point and the road surface. The ZMP stable region is a region set further inside the supporting polygon, and the ZMP can be accommodated in this region to make the aircraft highly stable.
[0143]
Then, a ZMP trajectory during the foot motion is set based on the foot motion and the ZMP stable area (step S3).
[0144]
In addition, for each part of the upper body (above the hip joint) of the body, a group is set such as a waist, a trunk, an upper limb, and a head (step S11). Then, a desired trajectory is set for each part group (step S12).
[0145]
Next, the group setting of each part is adjusted (re-grouping) (step S13), and priorities are given to these groups (step S14). The priority here refers to the order of input to the processing calculation for performing the attitude stabilization control of the airframe, and is assigned, for example, according to the mass operation amount. As a result, a desired trajectory group with priority for each part of the upper body of the aircraft is completed.
[0146]
In addition, a mass that can be used for moment compensation is calculated for each of the body groups of the upper body (step S15).
[0147]
Then, based on the foot motion, the ZMP trajectory, and the desired activation group for each upper body part group, the exercise pattern of each part group is input to the posture stabilization process in accordance with the priority set in step S14.
[0148]
In this posture stabilization processing, first, an initial value 1 is substituted for a processing variable i (step S20). Then, the moment amount on the target ZMP, that is, the total moment compensation amount at the time of setting the target trajectory for the i-th part group from the top is calculated (step S21). For a part for which the target trajectory has not been calculated, the desired trajectory is used.
[0149]
Next, the moment compensation amount is set using the mass that can be used for the moment compensation of the part calculated in step S15 (step S22), and the moment compensation amount is calculated (step S23).
[0150]
Next, the ZMP equation for the i-th part is derived using the calculated moment compensation amount for the i-th part (step S24), and the moment compensation movement for the part is calculated (step S25). It is possible to obtain the target trajectory for the i-th part from the top in the priority order.
[0151]
By performing such processing for all the part groups, a whole body movement pattern capable of stable movement (for example, walking) is generated.
[0152]
Since the acceleration sensors 96, 93, and 94 are provided on the waist and the left and right feet, respectively, the above ZMP balance equation is directly and accurately derived using the acceleration measurement results at these control target points. be able to. As a result, the posture stabilization control based on the ZMP stability determination criterion can be executed at high speed and more strictly according to the processing procedure shown in FIG.
[0153]
E. FIG. Falling operation of a legged mobile robot
As described in the preceding section D, the legged mobile robot 100 according to the present embodiment basically performs posture stabilization control during walking or other standing work based on the ZMP stability discrimination standard. To minimize the occurrence of a fall. However, if it becomes impossible to avoid a fall, a fall operation having an operation pattern that minimizes damage to the body is performed. For example, in the above-described ZMP balance equation, when an excessive external force F or an external force moment M is applied to the aircraft, the moment error component T cannot be canceled only by the aircraft operation, and the attitude cannot be maintained.
[0154]
FIG. 19 is a flowchart illustrating a schematic processing procedure of operation control of the body during legged work in the legged mobile robot 100 according to the present embodiment.
[0155]
During the operation of the aircraft, the ZMP balance equation ( ), And the ground contact position of the foot is always calculated (step S31).
[0156]
For example, when an external force is applied to the aircraft, it is determined whether or not the next foot contact position can be planned, that is, whether or not the moment error due to the external force can be eliminated by the action plan of the foot. (Step S32). Whether or not the installation position of the foot can be planned is determined in consideration of the movable angle of each joint of the leg, the torque of each joint actuator, joint force, angular velocity, angular acceleration, and the like. Of course, when an external force is applied, the moment error may be eliminated not only by the next step but also by several steps of leg steps.
[0157]
At this time, if planning of the foot is possible, walking and other legged movements are continued (step S33).
[0158]
On the other hand, when the planning of the foot becomes impossible because an excessive external force or an external force moment is applied to the body, the legged mobile robot 100 starts to fall (step S34).
[0159]
In the case of an upright walking type legged robot as shown in FIG. 1 and FIG. 2, since the position of the center of gravity is high, if the robot is accidentally dropped on the floor when falling, it is fatal to the robot itself or the opponent who collides due to falling. Risk of permanent damage. Therefore, in the present embodiment, a predetermined overturning operation is performed by changing the trajectory of the aircraft planned before the overturn to a posture that minimizes the ZMP support polygon. Basically, a falling motion is searched for based on the following two policies.
[0160]
(1) The variation ΔS / Δt of the area S of the support polygon of the body per time t is minimized.
(2) The support polygon when the floor falls is maximized.
[0161]
By adopting such a falling method, it is possible to minimize the damage by dispersing the impact received from the floor surface during the fall to the whole body.
[0162]
FIG. 20 shows, in the form of a flowchart, a processing procedure for the legged mobile robot 100 according to the present embodiment to perform the overturning operation due to the inability to plan the foot. The overturning operation is realized by synchronously and cooperatively driving the shoulder joint pitch axis 4, the trunk pitch axis 9, the hip joint pitch axis 12, and the knee joint pitch axis 14, which are connected in the height direction, in accordance with the basic policy described above. You. Such a processing procedure is actually realized by executing a predetermined machine operation control program in the main control unit 81 and controlling driving of each unit.
[0163]
First, a link that minimizes the change amount ΔS / Δt of the area S of the support polygon of the body per time t is searched (step S41).
[0164]
Next, a search is made for a target landing point of the link selected in step S41 that minimizes the amount of change ΔS / Δt (step S42).
[0165]
Then, the landing of the link selected in the preceding step at the target landing point depends on the limitations of the aircraft hardware (movable angle of each joint, torque of each joint actuator, joint force, angular velocity, angular acceleration, etc.). Then, it is determined whether or not execution is possible (step S43).
[0166]
If it is determined that it is impossible to land the link selected in the preceding step at the target landing point, the time variation Δt is incremented by a predetermined place (step S44), and then the step is performed. Returning to S41, reselection of the link and resetting of the target landing point of the link are performed.
[0167]
On the other hand, if the link selected in the preceding step can be landed on the target landing point, the selected link is landed on the target landing point (step S45).
[0168]
Next, it is determined whether or not the potential energy of the body is minimum, that is, whether or not the overturning operation has been completed (step S46).
[0169]
If the potential energy of the fuselage is not yet the minimum, the amount of change in time Δt is further increased by a predetermined land (step S47), and the next target landing point is set so as to enlarge the support polygon (step S47). S48).
[0170]
Then, the landing of the selected link at the target landing point can be executed due to constraints of the aircraft hardware (movable angles of each joint, torque of each joint actuator, joint force, angular velocity, angular acceleration, etc.). It is determined whether it is (Step S49).
[0171]
If it is determined that it is impossible to land the link selected in the preceding step at the target landing point, the process returns to step S41 to reselect the link and to set the target landing point of the link. Reset the settings.
[0172]
On the other hand, when it is possible to land the link selected in the preceding step at the target landing point, the process proceeds to step S45 to land the selected link at the target landing point.
[0173]
Then, when the potential energy of the aircraft is minimized (step S46), it means that the landing of the aircraft on the floor has been completed, and the entire processing routine ends.
[0174]
In FIG. 21, the legged mobile robot 100 is moved from a plurality of substantially parallel joint axes connected in the height direction such as the shoulder joint pitch axis 4, the trunk pitch axis 9, the hip joint pitch axis 12, and the knee joint pitch axis 14. A model in which the joint pitch axis is modeled as a link structure, and the joint pitch axes are driven in a synchronously emphasizing manner to fall over toward a supine posture is shown.
[0175]
It is assumed that the robot stands only at the sole that is the link end of the link structure (FIG. 21A).
[0176]
At this time, the application of the external force or the external force moment makes it impossible to cancel the moment error term T of the ZMP balancing equation, and the ZMP is supported in response to the ZMP deviating outside the ZMP stable region formed only by the sole. The falling motion is started while maintaining the polygon.
[0177]
In the overturning operation, first, a link that minimizes the amount of change ΔS / Δt per unit time t of the area S of the supporting polygon of the body is searched for, and the amount of change ΔS / Δt is minimized by a link including the hand. Search for the target landing point at hand. Then, whether landing of the selected link at the target landing point is feasible due to constraints of the aircraft hardware (movable angle of each joint, torque of each joint actuator, joint force, angular velocity, angular acceleration, etc.) Is determined.
[0178]
If the link can be executed on the hardware of the machine, another link will land in addition to the sole link that has already landed. Then, the ZMP is moved into the minimum support polygon formed by these landing links (FIG. 21 (2)).
[0179]
Next, as long as the machine hardware permits, the landing point is moved to expand the support polygon (FIG. 21 (3)).
[0180]
Then, when it is no longer possible to move the landing point due to the limitations of the body hardware such as the movable angle of each joint, the torque of each joint actuator, the joint force, angular velocity, angular acceleration, etc. It is determined whether it is possible to land on the leaving link sandwiched between the two.
[0181]
If it is possible to land the landing links between the landing links on the machine hardware, the landing links are landed to increase the number of landing links (FIG. 21D).
[0182]
Further, the landing point is moved and the support polygon is enlarged as far as the machine hardware permits (FIG. 21 (5)).
[0183]
Finally, one or more links from one end side of the link structure composed of a plurality of substantially parallel joint axes connected in the height direction and two or more links from the other end are separated from the floor, and With one or more of the links located at the middle, the ZMP within the supporting polygon is formed while the foot is further landed, and the posture in which the supporting polygon is maximized is formed. In this posture, if the potential energy of the body is minimum, the overturning operation is completed.
[0184]
FIG. 22 shows a state in which the actual machine falls from the standing posture to the supine posture. In the example shown in the figure, a state is shown in which the vehicle falls down in response to an external force from the front of the body, but the operation state can be grasped in four stages A to D.
[0185]
(A) Both legs supported before falling
This is a state in which the posture is controlled, and the actuators of all the joint parts are set to the frequency characteristics shown in FIG.
[0186]
In this case, the proportional gain of the servo controller is increased to increase the gain of the entire system so that the gain can be obtained up to a low frequency band. Further, as shown in the figure, the frequency characteristics are set so that the amount of phase lead is small in a high frequency range, and the stability can be guaranteed while not contributing much to quick response. Further, the viscous resistance of the motor is increased so as to be robust against disturbance such as vibration. As a result, the whole body is robust against disturbances such as vibrations with priority given to positioning accuracy.
[0187]
(B) An external force / shock leading to a fall is received and a fall operation is started to minimize the impact. An actuator of a joint part related to potential energy control such as an ankle roll axis, an ankle pitch axis, and a knee pitch axis is shown in FIG. Set to the frequency characteristics shown. In addition, the actuators of the other joints involved in the stable region control are maintained at the frequency characteristics shown in FIG.
[0188]
In a joint site related to potential energy control, such as an ankle roll axis, an ankle pitch axis, and a knee pitch axis, quick response and compliance are important. Therefore, by reducing the proportional gain and the gain of the entire system, the gain in the low frequency band is reduced and mechanical passivity (compliance) is easily obtained. Also, as shown in the figure, the amount of phase lead is increased in the middle and high frequency ranges so as to obtain quick response. Further, the viscous resistance of the motor is reduced so that mechanical passivity (compliance) can be easily obtained.
[0189]
(C) The sole leaves the floor, and the floor reaction force that had been requested on the sole becomes zero.
The actuator of the joint site related to the potential energy control such as the hip joint pitch axis is further set to the frequency characteristic shown in FIG. As a result, characteristics that prioritize mechanical passivity (compliance) and quick response are obtained at the joint site. In addition, the actuators of the other joints involved in the stable region control are maintained at the frequency characteristics shown in FIG.
[0190]
(D) The robot landed on the back and the overturning operation was completed
At the same time as the landing, the actuators of all the remaining joint parts are set to the frequency characteristics shown in FIG. As a result, characteristics that prioritize mechanical passivity (compliance) and quick response can be obtained as a whole body.
[0191]
By setting the characteristics of the actuator such that the low-frequency gain is large, the phase lead amount is small, and the viscous resistance of the joint is large as shown in FIG. 15, highly accurate positioning is possible. Therefore, in the states (B) and (C), the positioning accuracy of the joint, which is the main component for controlling the amount of ΔS / Δt, which is a reference when controlling the overturning operation, is secured, and the operation is stabilized. The nature increases.
[0192]
Further, by setting the characteristics of the actuator such that the low-frequency gain is small, the phase lead amount is large, and the viscous resistance of the joint is small as shown in FIG. And can give quick response. Accordingly, in the states (B) and (C), when the leg performs a passive operation such as partially utilizing the potential energy, the tracking control in a high band is performed while maintaining mechanical passivity (compliance). It becomes possible. Also, by giving mechanical passivity (compliance) at the moment of landing as in the state of (D), the entire robot becomes a soft system, so that the impact applied to the entire body can be reduced. .
[0193]
In FIG. 23, a plurality of substantially parallel joints in which the legged mobile robot 100 is connected in the height direction, such as the shoulder joint pitch axis 4, the trunk pitch axis 9, the hip joint pitch axis 12, and the knee joint pitch axis 14, are shown. The operation is modeled as a link structure composed of shafts, and the joint pitch axes are driven in a synchronously emphasized manner to fall toward a prone posture.
[0194]
It is assumed that the robot stands only at the sole that is the link end of the link structure (FIG. 23A).
[0195]
At this time, the application of the external force or the external force moment makes it impossible to cancel the moment error term T of the ZMP balancing equation, and the ZMP is supported in response to the ZMP deviating outside the ZMP stable region formed only by the sole. The falling motion is started while maintaining the polygon.
[0196]
In the overturning operation, first, a link that minimizes the amount of change ΔS / Δt per unit time t of the area S of the supporting polygon of the body is searched for, and the amount of change ΔS / Δt is minimized by a link including the hand. Search for the target landing point at hand. Then, whether landing of the selected link at the target landing point is feasible due to constraints of the aircraft hardware (movable angle of each joint, torque of each joint actuator, joint force, angular velocity, angular acceleration, etc.) Is determined.
[0197]
If the link can be executed on the hardware of the machine, another link will land in addition to the sole link that has already landed. Then, the ZMP is moved into the minimum support polygon formed by these landing links (FIG. 23 (2)).
[0198]
Next, as long as the aircraft hardware permits, the landing point is moved to enlarge the support polygon (FIG. 23 (3)).
[0199]
Then, when it is no longer possible to move the landing point due to the limitations of the body hardware such as the movable angle of each joint, the torque of each joint actuator, the joint force, angular velocity, angular acceleration, etc. It is determined whether it is possible to land on the leaving link sandwiched between the two.
[0200]
When it is possible to land the landing links between the landing links on the machine hardware, the landing links are landed to increase the number of landing links (FIG. 23 (4)).
[0201]
Further, the landing point is moved and the supporting polygon is enlarged as far as the machine hardware permits (FIG. 23 (5)).
[0202]
Finally, one or more links from one end side of the link structure composed of a plurality of substantially parallel joint axes connected in the height direction and two or more links from the other end are separated from the floor, and With one or more of the links located at the middle, the ZMP within the supporting polygon is formed while the foot is further landed, and the posture in which the supporting polygon is maximized is formed. In this posture, if the potential energy of the body is minimum, the overturning operation is completed.
[0203]
FIG. 24 shows a state in which the actual machine falls from the standing position to the prone position. In the example shown in the figure, a state is shown in which an overturning operation is performed by receiving an external force from the rear of the body, and the operation state can be grasped in four stages A to D.
[0204]
(A) Both legs supported before falling
This is a state in which the posture is controlled, and the actuators of all the joint parts are set to the frequency characteristics shown in FIG.
[0205]
In this case, the proportional gain of the servo controller is increased to increase the gain of the entire system so that the gain can be obtained up to a low frequency band. Further, as shown in the figure, the frequency characteristics are set so that the amount of phase lead is small in a high frequency range, and the stability can be guaranteed while not contributing much to quick response. Further, the viscous resistance of the motor is increased so as to be robust against disturbance such as vibration. As a result, the whole body is robust against disturbances such as vibrations with priority given to positioning accuracy.
[0206]
[0207]
(B) Start the overturning operation to minimize the impact by receiving the external force and impact leading to the overturn. Actuators of joints related to potential energy control such as ankle roll axis, ankle pitch axis, knee pitch axis, elbow pitch axis , And the frequency characteristics shown in FIG. In addition, the actuators of the other joints involved in the stable region control are maintained at the frequency characteristics shown in FIG.
[0208]
In a joint site related to potential energy control such as an ankle roll axis, an ankle pitch axis, a knee pitch axis, and an elbow pitch axis, quick response and compliance are important. Therefore, by reducing the proportional gain and the gain of the entire system, the gain in the low frequency band is reduced and mechanical passivity (compliance) is easily obtained. Also, as shown in the figure, the amount of phase lead is increased in the middle and high frequency ranges so as to obtain quick response. Further, the viscous resistance of the motor is reduced so that mechanical passivity (compliance) can be easily obtained.
[0209]
(C) The sole leaves the floor, and the floor reaction force that had been requested on the sole becomes zero.
Actuators at joints related to potential energy control, such as a hip joint pitch axis and a shoulder pitch axis, are further set to have frequency characteristics shown in FIG. As a result, characteristics that prioritize mechanical passivity (compliance) and quick response are obtained at the joint site. In addition, the actuators of the other joints involved in the stable region control are maintained at the frequency characteristics shown in FIG.
[0210]
(D) The robot landed on the back and the overturning operation was completed
At the same time as the landing, the actuators of all the remaining joint parts are set to the frequency characteristics shown in FIG. As a result, characteristics that prioritize mechanical passivity (compliance) and quick response can be obtained as a whole body.
[0211]
By setting the characteristics of the actuator such that the low-frequency gain is large, the phase lead amount is small, and the viscous resistance of the joint is large as shown in FIG. 15, highly accurate positioning is possible. Therefore, in the states (B) and (C), the positioning accuracy of the joint, which is the main component for controlling the amount of ΔS / Δt, which is a reference when controlling the overturning operation, is secured, and the operation is stabilized. The nature increases.
[0212]
Further, by setting the characteristics of the actuator such that the low-frequency gain is small, the phase lead amount is large, and the viscous resistance of the joint is small as shown in FIG. And can give quick response. Therefore, in the state (B), when the leg or the arm performs a passive operation in which the potential energy is half-utilized, a high-bandwidth tracking control is performed while maintaining mechanical passivity (compliance). Becomes possible. Further, in the state of (C), when the legs, arms, and the like operate while supporting the body even after landing, mechanical shock (compliance) is used to alleviate the impact applied to the body and reduce the impact. It is possible to perform a band following operation. Also, by giving mechanical passivity (compliance) at the moment of landing as in the state of (D), the entire robot becomes a soft system, so that the shock applied to the entire body can be reduced. .
[0213]
[Supplement]
The present invention has been described in detail with reference to the specific embodiments. However, it is obvious that those skilled in the art can modify or substitute the embodiment without departing from the spirit of the present invention.
[0214]
The gist of the present invention is not necessarily limited to products called “robots”. That is, as long as the mechanical device performs a motion similar to a human motion using an electric or magnetic action, the present invention similarly applies to a product belonging to another industrial field such as a toy. Can be applied.
[0215]
In this specification, a circuit example configured using a switching element formed of a bipolar transistor for switching control of a coil current to a motor coil has been described. It is obvious for a person skilled in the art that such a control circuit can be implemented.
[0216]
In short, the present invention has been disclosed by way of example, and the contents described in this specification should not be interpreted in a limited manner. In order to determine the gist of the present invention, the claims described at the beginning should be considered.
[0217]
【The invention's effect】
As described above in detail, according to the present invention, it is possible to reduce the damage to the robot as much as possible by controlling the motion of the entire body including the torso and the arms as well as the legs during the fall and fall. A legged mobile robot can be provided.
[0218]
Further, according to the present invention, a stable and highly efficient operation is achieved by dynamically or statically controlling two characteristics of the actuator itself and the characteristics of the controller of the actuator during the movement of the entire body during the fall and fall. An excellent legged mobile robot capable of realizing an operation can be provided.
[Brief description of the drawings]
FIG. 1 is a diagram showing a state in which a legged mobile robot used for carrying out the present invention is standing upright as viewed from the front.
FIG. 2 is a view showing a state in which the legged mobile robot provided for carrying out the present invention stands upright as viewed from behind.
FIG. 3 is a diagram schematically showing a configuration of a degree of freedom of a joint included in the legged mobile robot.
FIG. 4 is a diagram schematically showing a control system configuration of the legged mobile robot 100.
FIG. 5 is a diagram illustrating a configuration of a servo controller of an actuator.
FIG. 6 is a diagram illustrating frequency characteristics of a gain and a phase of a transfer function expression model of the motor and the speed reducer illustrated in FIG. 5;
7 is a diagram for explaining an example in which a phase compensation band is arbitrarily selected as a design example of phase compensation control in the servo controller shown in FIG. 5;
8 is a diagram for explaining an example in which the amount of phase compensation is arbitrarily selected as a design example of phase compensation control in the servo controller shown in FIG. 5;
9 is a diagram for explaining a design example of a controller that changes the magnitude of a series compensation gain indicated by K in the servo controller illustrated in FIG. 5;
FIG. 10 is a diagram showing an open-loop characteristic when a servo controller of an actuator is mounted so that a phase compensation amount is constant and a frequency band is arbitrarily selected as shown in FIG. 7;
11 is a diagram showing a state in which control of series compensation gain is further adopted in example C (s) -3 in which phase lead compensation is performed only in a high frequency band shown in FIG.
FIG. 12 is a diagram illustrating a configuration example of an equivalent circuit of a current control circuit for supplying a coil current of a DC motor to which a coil current control mechanism is applied.
FIG. 13 is a diagram showing a specific circuit configuration of an additional logic circuit.
FIG. 14 is a diagram showing output characteristics of each transistor control signal of the additional logic circuit when a BRAKE_PWM control signal having a predetermined duty ratio is input by PWM control, together with a coil current waveform characteristic and a torque output characteristic. .
FIG. 15 is a diagram showing frequency characteristics of an actuator that can be applied to a portion where positioning accuracy is important.
FIG. 16 is a diagram showing frequency characteristics of an actuator that can be applied to a portion where responsiveness and compliance are important.
17 is a diagram illustrating a multi-mass approximation model of the legged mobile robot 100 having the multi-joint degree of freedom configuration illustrated in FIG. 3;
FIG. 18 is a flowchart showing a processing procedure for generating a body motion capable of stably walking using the ZMP as a stability determination criterion in the legged mobile robot 100.
FIG. 19 is a flowchart showing a schematic processing procedure of operation control of the body of the legged mobile robot 100 during legged work.
FIG. 20 is a flowchart showing a processing procedure for the legged mobile robot 100 to perform a falling operation due to an inability to plan a foot.
FIG. 21 shows a legged mobile robot 100 including a plurality of substantially parallel joint axes connected in the height direction such as a shoulder joint pitch axis 4, a trunk pitch axis 9, a hip joint pitch axis 12, and a knee joint pitch axis 14. FIG. 10 is a diagram illustrating an operation of modeling as a link structure and driving each joint pitch axis in a synchronously emphasizing manner to fall toward a supine posture.
FIG. 22 is a diagram showing a state in which the actual machine falls from a standing posture to a supine posture.
FIG. 23 shows a legged mobile robot 100 including a plurality of substantially parallel joint axes connected in the height direction, such as a shoulder joint pitch axis 4, a trunk pitch axis 9, a hip joint pitch axis 12, and a knee joint pitch axis 14. It is a figure which modeled as a link structure, and showed the operation | movement which drives each joint pitch axis in a synchronous emphasis and falls down to a prone posture.
FIG. 24 is a diagram illustrating a state in which the actual machine falls from a standing posture to a prone posture.
[Explanation of symbols]
1 ... Yaw axis of neck joint
2A: First neck joint pitch axis
2B: Second neck joint (head) pitch axis
3 ... Neck joint roll axis
4: Shoulder joint pitch axis
5 ... shoulder joint roll axis
6 Upper arm yaw axis
7. Elbow joint pitch axis
8: Wrist joint yaw axis
9: trunk pitch axis
10 ... trunk roll axis
11 ... Yaw axis of hip joint
12: Hip joint pitch axis
13 ... hip joint roll axis
14. Knee joint pitch axis
15: Ankle joint pitch axis
16 ... Ankle joint roll axis
30 ... head unit, 40 ... trunk unit
50: arm unit, 51: upper arm unit
52: elbow joint unit, 53: forearm unit
60 ... leg unit, 61 ... thigh unit
62: knee joint unit, 63: shin unit
80: control unit, 81: main control unit
82: Peripheral circuit
91, 92 ... grounding confirmation sensor
93,94 ... Acceleration sensor
95 ... Attitude sensor
96 ... Acceleration sensor
100 ... Legged mobile robot

Claims (7)

A legged mobile robot comprising a plurality of joint parts including a plurality of movable legs, wherein a gain and a phase compensation control of an actuator servo controller and a control of a viscous resistance of an actuator motor are combined at each joint part. Means for controlling actuator characteristics to be performed;
Comprising a body motion control means for falling when controlling the body motion when falling,
The actuator characteristic control means includes a first actuator characteristic for increasing the low-frequency gain, a small amount of phase lead, and a high viscous resistance of the joint for each stage of the overturning operation. Switching between a second actuator characteristic that reduces the gain, increases the amount of phase lead, and reduces the viscous resistance of the joint;
A legged mobile robot, characterized in that: The falling aircraft operation control means,
Landing site searching means for searching for changes per unit time Δt of the area S of the supporting polygon formed by the ground contact point of the vehicle and the road surface;
Target landing at which the site selected by the landing site searching means should land so that the change ΔS / Δt per unit time Δt of the area S of the support polygon formed by the ground contact point of the vehicle and the road surface is minimized. Target landing point setting means for setting points,
Part landing means for landing the part selected by the landing part search means at the target landing point set by the target landing point setting means;
The legged mobile robot according to claim 1, wherein: The apparatus further includes a support polygon enlarging means for moving a landing part so as to further expand the newly formed support polygon by landing the selected part by the part landing means.
The legged mobile robot according to claim 2, wherein: Until the potential energy of the body is minimized and the overturning operation is completed, the landing operation of the part by the landing part searching means and the target landing point setting means and / or the support polygon by the support polygon enlarging means. Repeat the enlarging operation of the square,
The legged mobile robot according to claim 2, wherein: The actuator characteristic control means, for each stage of the overturning operation, the characteristics of the actuator of each joint site related to the stable area control, the low-frequency gain is large, the phase lead amount is small, the viscous resistance of the joint is maintained large,
The legged mobile robot according to claim 1, wherein: The actuator characteristic control means may include, for each stage of the overturning operation, a characteristic of the actuator at each joint site related to the potential energy control, wherein the low-frequency gain is large, the phase lead amount is small, and the viscous resistance of the joint is large. From the actuator characteristics of the above, switch to a second actuator characteristic that reduces the low-frequency gain, increases the amount of phase lead, and reduces the viscous resistance of the joint.
The legged mobile robot according to claim 1, wherein: The actuator characteristic control means switches the actuator characteristic of each joint site related to impact shock during landing to a second actuator characteristic that reduces the low-frequency gain, increases the amount of phase lead, and reduces the viscous resistance of the joint.
The legged mobile robot according to claim 1, wherein:

Images