JP2004086289A - Operation auxiliary device - Google Patents

Abstract

<P>PROBLEM TO BE SOLVED: To provide an operation auxiliary device for realizing smooth control. <P>SOLUTION: The operation auxiliary device is provided with a peripheral situation detection means 10 of a self-vehicle A, a self-vehicle state detection means 11, an evaluation function constitution means 6a constituting an evaluation function for calculating desirability of an operation, a controlled variable operation means 6b and an actuator 12 controlling a self-vehicle unit. The evaluation function constitution means 6a is provided with a label bestowal means 6-1 requesting bestowal of a label calculating a risk degree of an object being an obstacle, a label deletion means 6-2 requesting deletion of the label and a label management means 6-3, and it constitutes the evaluation function from an item for evaluating a size of a controlled variable of the self-vehicle A, an item for evaluating a self-vehicle state, an item for evaluating the risk degrees of the respective objects and a weighting coefficient on which significance of evaluation against the respective items is reflected. The controlled variable operation means 6b is provided with a weighting coefficient change means 6-4 changing a value of the weighting coefficient of the applied object when the label management means 6-3 receives the request of bestowal or deletion of the label. <P>COPYRIGHT: (C)2004,JPO

Description

のような関数を用いることができる。
【００２１】
自車両Ａに対する望ましい操作を算出するためには、評価関数として（１）式のような周囲環境に対する評価項だけでなく、自車両Ａに対する操作そのものを評価する項が必要である。例えば、自車両Ａに対する加減速の指令値をｕ_ｘとするなら、次のような評価項が必要になる。
【００２２】
【数２】

この他に、自車両Ａの状態を評価に加えることもできる。例えば、自車両Ａの希望走行車速をｖ^ｄに設定していたとすると、希望車速と実走行車速の偏差の二乗である
【００２３】
【数３】

を評価項に加えることができる。以上のような評価項に重み付け係数をつけて和をとった式を評価式として導入する。すなわち、
【００２４】
【数４】

で定義される関数Ｌを、ある瞬間における評価量として構成する。ここで、ｗ_Ｘ、ｗ_Ｖ、ｗ_ｉは、各評価項に対する重み付け係数である。評価関数は、関数Ｌを所定の評価区間にわたって積分した量として定義される。すなわち、
【００２５】
【数５】

が評価関数である。ここで、ｔは現在の時刻を表し、Ｔは評価区間の長さを表す。評価関数構成手段６ａでは、以上のような処理が行なわれ、構成された評価関数に関する情報が制御量演算手段６ｂに送られる。
【００２６】
制御量演算手段６ｂは、評価関数設定手段６ａで設定された評価関数の情報を受けて、アクチュエータを制御する制御量の算出を行なう。制御量の演算は、評価関数（５）式が最小となるような自車両Ａに対する操作量ｕ_ｘ（τ）、ｔ≦τ≦ｔ＋Ｔを求め、算出された操作量ｕ_ｘ（τ）に沿うような操作を運転者に促すように反力を制御する指令値を反力モーター５に送る。
【００２７】
評価関数（５）式を評価するためには、評価区間ｔ≦τ≦ｔ＋Ｔにおける自車両Ａおよび他物体の状態を予測する必要があるが、図３の場面では、予測式として次のようなものを考えることができる。
自車両Ａに関しては、
【００２８】
【数６】

で状態量を予測する。後方車Ｃの予測式としては、自車両Ａに追従走行しているモデルとして、
【００２９】
【数７】

を用いることができる。ここで、ｋ_１、ｋ_２は後方車Ｃの追従特性を表す定数、ｈ_Ｂは後方車Ｃの目標車間時間である。先行車Ｂについては、一定速度で走行するものとして、
【００３０】
【数８】

というモデルを用いる。
【００３１】
以上のように、評価区間ｔ≦τ≦ｔ＋Ｔの間の自車両Ａおよび各物体の状態量を予測する式を定義できれば、（５）式を最小にするような制御量を求める問題は、最適制御問題として定式化されており、その系統的な解法はよく研究されている（例えば、文献１：加藤寛一郎著　工学的最適制御　非線形へのアプローチ、など）。
【００３２】
ここで問題となるのは、先行車Ｂがラベル付与区間内に進入し、評価項（４）式右辺第三項において、それまでの後方車Ｃだけの評価項に加えて、先行車Ｂに対する評価項が加わった場合である。一般に、最適制御問題は、連続な予測式と連続な評価関数を用いていれば、評価の始端における状態量の連続的な変化に対して、最適解も連続的に変化する。最適解が連続的に変化すれば、最適解から生成されるアクチュエータ１２の制御量も連続的に変化するので、提示される反力も滑らかなものになる。ところが、先行車Ｂに対する評価項が加わった場合には、評価関数自体が不連続に変化しているので、得られる最適解もそれを反映した不連続性が現れ、提示される反力も不連続に変化する。また、文献２：Ｔ．　Ｏｈｔｓｕｋａ，　“Ｃｏｎｔｉｎｕａｔｉｏｎ／ＧＭＲＥＳ　ｍｅｔｈｏｄ　ｆｏｒ　ｆａｓｔ　ａｌｇｏｒｉｔｈｍ　ｏｆ　ｎｏｎｌｉｎｅａｒ　ｒｅｃｅｄｉｎｇｈｏｒｉｚｏｎ　ｃｏｎｔｒｏｌ”　Ｐｒｏｃ．　３９ｔｈ　ＩＥＥＥ　Ｃｏｎｆｅｒｅｎｃｅ　ｏｎ　Ｄｅｃｉｓｉｏｎ　ａｎｄ　Ｃｏｎｔｒｏｌ，　ｐｐ．７６６−７７１，　２０００．に示されているような、解の連続性を前提に最適制御問題を高速に解くアルゴリズムを利用する場合にも、評価関数の不連続性は解の連続性の仮定に反するものであり、アルゴリズムの性能劣化の懸念が生じる。
【００３３】
このような不連続性を補償するために、制御量演算手段６ｂに重み付け係数変更手段６−４を設ける。先行車Ｂにラベルが付与される前は、評価式（４）は
【００３４】
【数９】

であり、先行車Ｂにラベルが付与されると、
【００３５】
【数１０】

に変化する。このままだと、前述の通り、評価項ｗ_２ｌ_２が加わることによる不連続性の問題点が現れてくる恐れがある。そこで、本来は固定のパラメータである重み付け係数ｗ_２を時刻ｔとともに変化する時変のパラメータとする。すなわち、
【００３６】
【数１１】

とする。今、先行車Ｂにラベルが付与された時刻をｔ_０として、ｗ_２（ｔ_０）＝０となるようにｗ_２（ｔ）を設定すれば、時刻ｔ_０において評価関数（１１）式はＣにラベルが付与される前の評価関数と一致するので、評価関数の連続性を保つことができる。時刻ｔ_０以降は、ｗ_２（ｔ）を本来の設計値であるｗ_２まで漸増させていけば、（９）式から（１０）式まで連続的に評価関数を変形していくことができるので、結果として生成される最適解も連続的に変化していき、制御量も滑らかな変化をするようになる。
【００３７】
ｗ_２（ｔ）の変化のさせ方としては、例えば図４のように、変化率を一定に保って、本来の設計値であるｗ_２まで漸増させる方法が考えられる。式で表現すると次式のようになる。
【００３８】
【数１２】

ただし、Ｔ_ｔはｗ_２（ｔ）の変化率を決定するパラメータである。一般にＴ_ｔを大きくとれば、評価関数の変化はゆっくりとしたものとなり、得られる制御量もより滑らかに変化するものと期待できる。しかし、Ｔ_ｔを長くとり過ぎると、新たに検出した物体に対する評価が反映されにくくなり、新たに検出した物体が自車両Ａに対して急接近してくるような場面では、対応が遅れ気味となり、かえって運転者に違和感を与える恐れがある。そこで、パラメータＴ_ｔを、新たに検出した物体の自車両Ａに対する衝突時間Ｔ_Ｃに応じて調節するという方法が考えられる。すなわち、図３の場面においては、
【００３９】
【数１３】

【００４０】
【数１４】

のような調整方法を用いることができる。ここで、Ｔ_ｔ ^ｍｉｎ、Ｔ_ｔ ^ｍａｘはパラメータＴ_ｔの調節範囲の下限と上限を表す定数であり、Ｔ_Ｃ ^ｍｉｎ、Ｔ_Ｃ ^ｍａｘは時間の次元を持つ適当な定数である。
【００４１】
具体的な反力指令値の演算方法としては、例えば以下のようなものが考えられる。評価関数（５）式を最小にする最適解の現在時刻ｔにおける値をｕ_ｘ ^＊（ｔ）とし、車両の加速度がｕ_ｘ ^＊（ｔ）となるようなアクセルペダル３の踏み込み角をθ^＊（ｔ）とする。実際に運転者が操作しているアクセルペダル３の踏み込み角をθ（ｔ）とし、さらに反力制御を適用しない通常のアクセルペダル３の反力特性を、Ｆ（θ）で表す。このとき、反力モータ−の駆動力によって、実際の反力が次式と一致するように制御指令を出力する。
【００４２】
【数１５】

ただし、Ｋは適当な正のゲインであり、ｆは反力修正量の上限値であり、
【００４３】
【数１６】

である。すなわち、実際の踏み込み角θ（ｔ）が最適解に対応する踏み込み角よりも大きい場合には、反力が大きくなって運転者に減速を促し、逆の場合には、反力が小さくなって、まだアクセルペダル３を踏み込む余裕があることを知らせる。
【００４４】
以上の処理をフローチャートにまとめたものが図５である。
ステップ１において周囲状況検出手段１０からの信号を受けて、障害物となり得る物体を識別する。
ステップ２において、新たにラベル付与区間内に入ってきた障害物の有無をチェックする。新たな障害物を検出した場合には、それに対応する評価項と予測式をそれぞれステップ３とステップ４で生成した上で、ステップ５へ進む。
ステップ５では最適制御問題を解く際に用いる重み付け係数の設定を行なう。既に本来の設計値まで到達している重み付け係数はそのままにして、重み付け係数が漸増中のものは新たな値に更新し、新しく追加された重み付け係数は０に設定する。
ステップ６では、そのように設定された重み付け係数に基づいて、最適制御問題を解いて最適解を得る。
ステップ７では、得られた最適解の始端点の値をもとに（１５）式に従って反力の指令値を計算して、１サイクルの処理を終了する。
【００４５】
図６に本発明を図３の場面に適用した場合に得られる結果の例を示す。
簡単のため、ここでは運転者は算出された最適解ｕ_ｘ ^＊（ｔ）を正確にトレースしているという場合を考えて、ｕ_ｘ ^＊（ｔ）の値の変化を示す。先行車Ｂを検出する前は、希望車速ｖ^ｄで一定速走行をしていたものとしている。図６に示したように、先行車Ｂに対する評価を重み付け係数を本来の設計値に固定したままで導入した場合には、先行車Ｂに対する評価を入れた瞬間に最適解が不連続的に変化している。最適解の不連続的な変化は（１５）式の反力演算式を通して提示される反力の不連続性の原因となる。一方、本発明の第一の実施の形態で示したように重み付け係数の調整を行なった場合には、最適解が連続的に変化しているため、提示される反力も連続的に変化することになる。
【００４６】
以上のように、本発明の第一の実施の形態は、自車両Ａの周囲状況を検出する周囲状況検出手段１０と、自車両Ａの状態を検出する自車両状態検出手段１１と、周囲状況検出手段１０と自車両状態検出手段１１からの信号により自車両Ａに対する運転操作の望ましさを算出する評価関数を構成する評価関数構成手段６ａと、運転者に望ましい操作を促すために自車両機器に対する制御信号を演算する制御量演算手段６ｂと、制御量演算手段６ｂからの信号により自車両機器を制御するアクチュエータ１２とを有し、評価関数構成手段６ａは、周囲状況検出手段１０からの信号により自車両Ａに対して障害となり得る物体に、物体に対するリスク度を算出するためのラベルを付与するように要求するラベル付与手段６−１と、ラベルを付与された物体に対するリスク度評価を打ち切る場合にラベルを削除するように要求するラベル削除手段６−２と、ラベルとラベルが付与された物体との対応関係を記録し、ラベル付与手段６−１とラベル削除手段６−２からの要求に応じて対応関係を更新するラベル管理手段６−３とを備え、自車両Ａに対する操作量の大きさを評価する項と、自車両Ａの状態を評価する項と、ラベルを付与された各物体の自車両Ａに対するリスク度を評価する項と、各項に対する評価の重要度を反映する重み付け係数とから評価関数を構成し、制御量演算手段６ｂは、ラベル管理手段６−３がラベルの付与または削除の要求を受け取った場合に、該当する物体に対する重み付け係数の値を変更する重み付け係数変更手段６−４を備えている。このような構成により、評価関数の不連続な変化を、重み付け係数の調整によって補償することができるので、算出される制御指令を滑らかなものにすることができる。
【００４７】
なお、評価関数による定式化をした場合、障害物に対応する評価項とそれに対する重みが設定されるが、そこで設定される重みに物理的な意味はなく、運転者が障害物との接近をどの程度許容するかによって最適な値が定まる主観的なパラメータなので、本実施例にあっては推定パラメータでなく、制御パラメータとみなして演算処理している。また、運転操作補助の際考慮しなければならない自車両周囲の障害物は必ずしも一つとは限らず、走行する環境によって障害物の数が時々刻々と変化しており、障害物数の変化に伴って、評価関数の構成自体も変化している。一方、運転操作補助の制御において、運転者に違和感を与えないために滑らかな制御が求められることが、評価関数が不連続に変化してしまうと、算出される制御指令も不連続に変化してしまうので、必ずしも滑らかな制御にならない。また、一般に評価関数を最小にする操作量を求める最適化演算は、計算負荷が高くなることが多く、特に制御対象や評価関数が非線形の場合には計算時間が大きな問題になる。
【００４８】
このため、本実施例では、上記構成をとることにより解の変化の連続性を仮定することで計算時間を短縮し、評価関数の不連続な変化や、解の変化の連続性が外れることがないようにしており、これによりアルゴリズムの性能に悪影響を与えることをなくし、滑らかな違和感ない制御を実現している。
【００４９】
また、ラベル付与手段６−１は、自車両Ａに対して障害となり得る物体と自車両Ａとの距離が所定値よりも小さくなった場合に、対象となる物体にラベルを付与するように要求し、ラベル削除手段６−２は、物体と自車両Ａとの距離が別の所定値よりも大きくなったか、あるいは周囲状況検出手段１０で検出できなくなった場合に、ラベルの削除を要求するようになっている。このような構成により、評価対象となる障害物の判定にヒステリシスをつけることができるようになるので、障害物が評価対象となる範囲の境界上に留まるような場合においても、ラベルの付与と削除が頻繁に繰り返されるような事態を防止できるようになり、より安定した制御指令を算出することができる。
【００５０】
また、制御量演算手段６ｂは、ラベル管理手段６−３が新たなラベル付与の要求を検出した場合に、新たにラベルが割り当てられた物体のリスク度に対する重み付け係数の値を０にセットし、その後、所定時間にわたって重み付け係数の値を漸増させるようになっている。このような構成により、新たな障害物を評価に入れる必要性が生じた瞬間において、評価関数の変化の連続性を保ちながら評価に組み込んでいくようになっているので、新たな障害物の検出に伴う制御量の変化を滑らかなものにすることができる。
【００５１】
また、式１２、１３に示したように、重み付け係数変更手段における重み付け係数の値の変化率を一定値とし、値の変化率を新たに検出された物体に対する衝突時間（＝車間距離／相対速度）に応じて決めるようになっている。このような構成により、重み付け係数の変化率を障害物との衝突時間に応じて決めているので、物体の接近速度に応じて制御系の応答の速さと制御の滑らかさのバランスを変えることができる。
【００５２】
また、評価関数構成手段６ａは、評価対象となる自車両Ａに対する操作量の大きさとして、自車両Ａに対する加減速指令を評価に含むような評価関数を構成するようになっている。このような構成により、自車両Ａに対する操作量を評価対象としているので、過大な制御量が算出されることを防止し、現実的な操作の範囲内で制御量を算出することができる。
【００５３】
また、評価関数構成手段６ａは、評価対象となる自車両Ａの状態として、自車両Ａの速度を評価に含むような評価関数を構成するようになっている。このような構成により、自車両Ａに対する状態を評価対象としているので、運転者の希望走行車速を考慮した制御量を算出し、運転者の意向に沿った運転操作の補助を行なうことができる。
【００５４】
また、評価関数構成手段６ａは、ラベルを付与された物体の自車両Ａに対するリスク度として、自車両Ａに対する距離が小さいほど、あるいは自車両Ａに接近する速度が大きければ大きいほど、リスク度を大きく評価する項を用いて評価関数を構成するようになっている。このような構成により、リスク度の評価を障害物までの距離と相対速度に応じて決めているので、自車両Ａが障害物に衝突する潜在的なリスク度合いを客観的に評価して、障害物と適切な距離を保つことを促すような制御量を算出することができる。
【００５５】
なお、重み付け係数変更手段６−４における重み付け係数の値の変化率を、値を変化させる区間の始端と終端において０に設定し、値の平均変化率を新たに検出された物体に対する衝突時間に応じて決めることで、重み付け係数の変化率を連続になるように構成し、重み付け係数の値が連続となっている場合よりもさらに滑らかな制御を実現することもできる。
【００５６】
また、重み付け係数変更手段６−４における重み付け係数の値の変化率を一定値とし、値の変化率を該当する物体に対する相対速度に応じて決めることで、重み付け係数の変化率を障害物との相対速度に応じて決め、制御量演算の際に考慮する必要性が低い物体を早く評価対象から外すこともできる。
【００５７】
更に、重み付け係数変更手段６−４における重み付け係数の値の変化率を、値を変化させる区間の始端と終端において０に設定し、値の平均変化率を削除対象となった物体に対する相対速度に応じて決めることで、重み付け係数の変化率も連続になるように構成し、重み付け係数の値が連続となっている場合よりもさらに滑らかな制御を実現することができる。
【００５８】
以上が本発明の第一の実施の形態である。
【００５９】
第二の実施の形態
本発明の第二の実施の形態を図７から図１０までの図面に基づいて説明する。
第二の実施の形態の主な構成は、第一の実施の形態と同一であり、配置図は図１、構成図は図２で表される。
以下、図７に示すような道路状況を例にとり、第二の実施の形態における各ブロックの具体的な動作を説明する。
【００６０】
図７では、片側一車線の道路を先行車Ｂ、自車両Ａ、後方車Ｃの順に走行している状態から、後方車Ｃの車速が低下して、後方車Ｃが自車両Ａの障害物評価区間であるラベル保持区間の外側に出た場面を示している。なお、後方車Ｃにはラベル“１”が、先行車Ｃにはラベル“２”が付与されているものとする。第一の実施の形態と同じ評価式を用いているとすると、評価式は（１０）式と同じになる。
【００６１】
後方車Ｃがラベル保持区間の外側に出たことで、ラベル削除手段６−２は、ラベル管理手段６−３に対して、後方車Ｃに付与されたラベル“１”を削除するように要求を出す。削除要求に即座に応えた場合、評価式は
【００６２】
【数１７】

のようになり、後方車Ｃに対する評価項ｗ_１ｌ_１の分だけ不連続性が生じ、第一の実施の形態で説明したことと同様の不都合が生じる恐れがある。そこで、ラベル管理手段６−３は削除要求を受け取っても、ただちに後方車Ｃのラベル“１”を削除せず、制御量演算手段６ｂにラベル“１”に対応する重み付け係数を変更するように指令する。すなわち、評価式を
【００６３】
【数１８】

として、重み付け係数ｗ_１（ｔ）を本来の設計値ｗ_１から漸減させ、値が０まで減少したところで、ラベル“１”を削除する。そうすることにより、評価関数の連続性を保つことができ、生成される制御量の変化を滑らかなものにすることができる。
【００６４】
重み付け係数ｗ_１（ｔ）の変化のさせ方としては、第一の実施の形態の（１２）式に示したように、変化率を一定に保って漸減させる方法をとることもできるが、ラベルを削除する場面では削除対象となる物体に対する応答の速さがあまり問題にならないので、より滑らかさを重視した重み付け係数の変化方法をとることができる。具体的には、重み付け係数を単に連続的に変化させるだけではなく、図８に示したように変化率も連続となるように滑らかに変化させる。式の形で書くと、例えば、
【００６５】
【数１９】

のような式が考えられる。ここでｔ_０はラベル管理手段６−３がラベル“１”の削除要求を受け取った時刻である。（１９）式のような変化をさせると、
【００６６】
【数２０】

が成り立ち、重み付け係数を変化させる区間の両端でその変化率が０になり、値だけでなく変化率も連続となるので、より滑らかな変化にすることができる。このような滑らかさを重視した変化式は、変化の初期における値の変化を抑制することになるので、応答性は遅くなるが、応答性があまり重要でない場面では問題とはならない。平均的な応答速度については、第一の実施の形態と同様に、パラメータＴ_ｔを変えることで調節することが可能である。ただ、ラベルを削除する場面は、該当物体が離れていく場面であり、衝突時間が無限大となってしまうので、（１３）式のような衝突時間を基準にした調節方法は意味をなさない。そこで、衝突時間ではなく、相対速度を基準に調節する次式のような方法が考えられる。
【００６７】
【数２１】

ただし、Ｒ_ｄ ^ｍａｘ、Ｒ_ｄ ^ｍｉｎは速度の次元を持つ適当な定数である。
【００６８】
物体が離れていく場面では、重み付け係数が０になる前に、該当物体がセンサの検出範囲外に出てしまい、評価項（１）式が計算できなくなってしまう事態が考えられる。特に、ラベル保持区間をセンサの検出範囲と同じ範囲にとった場合には、該当物体のロストがラベル削除要求のトリガーとなるので、該当物体に関する情報がそれ以上得られない。そのような場合には、該当物体をロストする直前のデータを用いて、ロストした物体の位置と速度を推定し、推定値を用いて（１）式の評価を行なう。例えば、図７の場面で、後方車Ｃをロストする直前の時刻ｔ^ｂｌにおける位置と速度が、それぞれｘ_Ｂ ^ｂｌ、ｖ_Ｂ ^ｂｌであったとき、次式に従って推定値を生成する。
【００６９】
【数２２】

以上の処理をフローチャートにまとめたものが図９である。
ステップ１において、周囲状況検出手段１０からの信号を受けて、障害物となり得るものを識別する。
ステップ２において、ラベル保持区間外へ出た障害物の有無をチェックする。
該当する物体がある場合には、ステップ３でその物体に対してラベル削除要求を出しておく。ステップ３の段階では、削除要求が出されるだけで、ただちに削除はしない。
ステップ４では、ラベルが付与されていた物体の中で、センサがロストした物体の有無をチェックする。
もしあった場合には、ステップ５で（２２）式のような推定値を生成し、推定値を検出値の代わりに利用する。
ステップ６では重み付け係数の更新を行なう。ラベル削除要求が出されている物体に対する重みは減少させ、そうでない物体の重みは保持する。
ステップ７では、更新された重み付け係数の値をチェックし、値が０になったものの有無を調べる。
値が０になったものがある場合には、ステップ８で該当するラベル、および評価項と予測式を削除する。
ステップ９では、残った評価項と予測式から構成された評価関数と予測式を用いて最適制御問題を解き、最適解を得る。
ステップ１０では、得られた最適解の始端点の値をもとに（１５）式に従って反力の指令値を計算して、１サイクルの処理を終了する。
【００７０】
図１０に本発明を図７の場面に適用した場合に得られる結果の例を示す。
第一の実施の形態と同様に、ここでは運転者は算出された最適解ｕ_ｘ ^＊（ｔ）を正確にトレースしているという場合を考えて、ｕ_ｘ ^＊（ｔ）の値の変化を示す。時刻０から時刻ｔ^ｂｄまでは三車が同一の速度で一定速走行していたとし、時刻ｔ^ｂｄから後方車Ｃが減速を始めて、時刻ｔ_０で後方車Ｃがラベル保持区間の外側に出た場面を想定している。後方車Ｃとの車間距離が開くと、先行車Ｃとの車間距離を広げるため、最適解は若干の減速を指令した後、再度加速するような解となる。図１０には、重み付け係数を固定とした場合と、重み付け係数を（１２）式と同様に一定の変化率で漸減させた場合と、重み付け係数を（１９）式のように変化率にも連続性をもたせて漸減させた場合の３通りの結果を示している。重み係数を固定した場合には、第一の実施の形態と同様に、時刻ｔ_０でｕ_ｘ ^＊（ｔ）の値が不連続に変化してしまう。一方、変化率を一定にして重みを漸減させた場合には、時刻ｔ_０におけるｕ_ｘ ^＊（ｔ）の連続性は保たれているが、変化率に若干のギャップが見られる。重み付け係数の変化率にも連続性を持たせた場合には、ｕ_ｘ ^＊（ｔ）も滑らかに変化している。
【００７１】
以上のように、本発明の第二の実施の形態は、制御量演算手段６ｂは、ラベル管理手段６−３がラベルの削除要求を検出した場合に、削除対象となったラベルが割り当てられた物体のリスク度に対する重み付け係数の値を所定時間にわたって０まで漸減させ、所定時間後に該当するラベルを削除するようになっている。このような構成により、検出物体を制御量算出の評価から外す時に、評価関数の変化の連続性を保ちながら評価から外す構成としているので、リスク度評価対象物体の変化に伴う制御量の変化を滑らかなものにすることができる。
【００７２】
また、式２１、２２に示したように、制御量演算手段６ｂは、重み付け係数の値が０になる前に対象となる物体が周囲状況検出手段１０で検出できなくなった場合に、検出できなくなった直前の時点における物体との距離と相対速度から外挿した暫定値を用いて制御量の演算を行なうようになっている。このような構成により、周囲状況検出手段１０から直接検出されるデータだけでなく、必要に応じて暫定値を用いているので、障害物のロストによる制御量演算への影響を小さくすることができる。
【００７３】
以上が本発明の第二の実施の形態である。
【００７４】
第三の実施の形態
本発明の第三の実施の形態を図１１および図１２の図面に基づいて説明する。
第三の実施の形態の主な構成は、第一の実施の形態と同一であり、配置図は図１、構成図は図２で表される。
以下、図１１に示すような道路状況を例にとり、第三の実施の形態における各ブロックの具体的な動作を説明する。
【００７５】
図１１では、片側一車線の道路を先行車Ｂ、自車両Ａ、後方車Ｃの順にほぼ同じ速度で走行している状態を示している。自車両Ａは先行車Ｂにやや接近して走行している。先行車Ｂ、後方車Ｃともに、自車両Ａが障害物としての評価を行なうラベル付与区間の内部を走行しているものとする。ここでは、そのような場面において、運転操作補助装置のシステム起動スイッチが押されて、運転補助の動作が開始される状況を特に考えることにする。起動スイッチが押された瞬間を時刻ｔ＝０と定義しておく。
【００７６】
システムが起動されると、周囲状況検出手段１０は先行車Ｂと後方車Ｃをラベル付与区間の内部に検出する。その結果が演算部６に送られると、ラベル付与手段６−１はただちにラベル管理手段６−３に対して、両車にラベルを付与するように要求する。ラベル管理手段６−３は要求を受けて、各車に適当なラベルを付与する。ここでは、先行車Ｂにラベル“１”が、後方車Ｃにラベル“２”が付与されたものとする。
【００７７】
評価関数構成手段６ａは、検出結果を受けて（１０）式と同じ評価式を構成し、制御量演算手段６ｂへ処理をわたす。制御量演算手段６ｂでは、重み付け係数の修正が行なわれるが、この場合、ラベル１、２両方とも新規検出対象であるので、両方の重み付け係数が可変となるように評価式が再構成される。すなわち、
【００７８】
【数２３】

と評価式が設定される。重み付け係数の変化のさせ方については、第一の実施の形態における（１２）〜（１４）式に準じた方法を利用することができる。また、各車の挙動予測の式についても、第一の実施の形態における（６）〜（８）式を流用することができる。
【００７９】
システム起動直後の動作が他の時点と異なるのは、システム起動直後はそれ以前に生成された最適解に関する情報を利用できないことである。このことは、最適化問題を解く際に、直前に得られた解を利用して効率良く問題を解く手法が利用できなくなり、計算負荷が高くなるという問題と、最初に得られる解の値によっては提示される反力が不連続に変化して、乗員に違和感を与えるという問題がある。
【００８０】
前記文献２には、このような問題を解決する方法として、評価関数（５）式における評価区間の長さＴを可変とし、時刻ｔ＝０において評価区間の長さを０として、その後、時間の経過とともに評価区間の長さを本来の設計値まで漸増させていくという方法が開示されている。評価区間の長さを０にした場合には、最適解は常に操作量が０となる自明解が得られる。自明解を得るには実質的に計算が不要なので、計算負荷の問題が回避される。また、得られる操作量が０なので、システム起動直後は反力を修正するような指令が発せられることがなくなり、不連続性の問題の解決にもなっている。
【００８１】
評価区間の長さの設定方法としては、次式のようなものが例として挙げられる。
【００８２】
【数２４】

ただし、Ｔ（ｔ）は時刻ｔにおける評価区間の長さ、Ｔ_０は本来の評価区間長の設計値、αは適当な正の定数である。
【００８３】
以上のような評価区間長の調整と重み付け係数の調整を併用することによって、システム起動から定常運転に至るまでの制御量の変化を滑らかなものにすることができる。
【００８４】
図１２に図１１の場面でシステムを起動した直後に得られる結果の例を示す。
第一の実施の形態と同様、ここでもｕ_ｘ ^＊（ｔ）の値の変化を示す。また、評価区間長と重み付け係数を両方とも固定とした場合、評価区間長は可変とするが、重み付け係数は固定とした場合、評価区間長、重み付け係数ともに可変とした場合の３通りの結果を示している。
【００８５】
図１１の場面では、自車両Ａが先行車Ｂに近づき過ぎているので、車間距離を広げるために減速を促す指令が出力されるはずである。評価区間と重み付け係数を両方とも固定とした場合には、システム起動直後に減速指令が発生し、大きな補助反力が発生する。評価区間を可変とした場合には、初期最適解は０となり、時刻ｔ＝０において補助反力は発生しない。重み付け係数を固定とした場合には時刻ｔ＝０における最適解の変化率に不連続性が残るが、重み付け係数も可変とした場合には、最適解の変化率も連続となり、滑らかな反力提示を行なうことができる。
【００８６】
以上が本発明の第三の実施の形態である。
【００８７】
以上本発明を実施の形態に基づいて具体的に説明したが、本発明は上記実施の形態に限定されるものではなく、その要旨を逸脱しない範囲において種々変更可能であることは勿論である。
【図面の簡単な説明】
【図１】本発明の第一の実施の形態の一配置図を示す図である。
【図２】本発明の第一の実施の形態の構成図を示す図である。
【図３】本発明の第一の実施の形態の適用場面の一例を示す図である。
【図４】本発明の第一の実施の形態の重み付け係数の変化方法の一例を示す図である。
【図５】本発明の第一の実施の形態における処理の手順を示すフローチャートである。
【図６】本発明の第一の実施の形態を適用した場合に得られる結果の一例を示す図である。
【図７】本発明の第二の実施の形態の適用場面の一例を示す図である。
【図８】本発明の第二の実施の形態の重み付け係数の変化方法の一例を示す図である。
【図９】本発明の第二の実施の形態における処理の手順を示すフローチャートである。
【図１０】本発明の第二の実施の形態を適用した場合に得られる結果の一例を示す図である。
【図１１】本発明の第三の実施の形態の適用場面の一例を示す図である。
【図１２】本発明の第三の実施の形態を適用した場合に得られる結果の一例を示す図である。
【符号の説明】
１ａ…前方レーダー
１ｂ…画像センサ
１ｃ…後方レーダー
１ｄ…側方センサ
２…車速センサ
３…アクセルペダル
４…アクセルペダル角度センサ
５…反力モーター
６…演算部
６ａ…評価関数構成手段
６−１…ラベル付与手段
６−２…ラベル削除手段
６−３…ラベル管理手段
６ｂ…制御量演算手段
６−４…重み付け係数変更手段
７…システム起動／停止スイッチ
Ａ…自車両
１０…周囲状況検出手段
１１…自車両状態検出手段
１２…アクチュエータ[0001]
TECHNICAL FIELD OF THE INVENTION
The present invention relates to a driving operation assisting device for assisting a driving operation of a vehicle.
[0002]
[Prior art]
In general, in the design of a control law, an evaluation function for evaluating control performance is often defined, and a solution obtained by solving a problem of finding an operation amount that minimizes the evaluation function may be used. In many cases, the control performance is evaluated in such a manner that some weighting is performed and a desired control characteristic can be obtained by adjusting the weighting.
In particular, when the operating conditions of the control system greatly change, online adjustment of the weights according to the operating conditions is an essential technique for maintaining the performance of the control system. JP-A-6-130970 is an example of such a technique. In Japanese Patent Application Laid-Open No. Hei 6-130970, the weighting coefficient (filter coefficient) is a parameter expressing the characteristics of the sound transmission system and the noise generation state. Adjustment of weighting coefficients corresponds to estimation of their physical characteristics, and a technique relating to adjustment of weighting coefficients based on the LMS algorithm is disclosed.
[0003]
[Problems to be solved by the invention]
The weighting factor does not always reflect a physical characteristic. In particular, when designing a control system of a device that assists a driver in driving, a parameter reflecting a subjective factor such as a driver's feeling may be used as a weighting coefficient. In this case, since the adjustment of the weighting coefficient has a different meaning from the estimation of the physical characteristics, it does not make sense to use the formalization or the adjustment technique of the weighting coefficient as in the related art as it is.
[0004]
An object of the present invention is to provide a driving assist system that can realize smooth control that does not give a driver an uncomfortable feeling.
[0005]
[Means for Solving the Problems]
In order to solve the above problems, a driving operation assisting device according to the present invention includes: a surrounding state detecting unit that detects a surrounding state of a host vehicle; a host vehicle state detecting unit that detects a state of the host vehicle; Evaluation function configuration means for configuring an evaluation function for calculating desirability; control amount calculation means for calculating a control signal for the own vehicle device; and an actuator for controlling the own vehicle device. A labeling means for requesting labeling to calculate a risk degree for an object that may be an obstacle to a vehicle; a label deleting means for requesting deletion of a label when the risk degree evaluation for the labeling object is terminated; and a label. And label management means for recording the correspondence between the label and the object to which the label is attached, and updating the correspondence in response to a request from the label assignment means and the label deletion means. Reflects the importance of the evaluation of the terms for evaluating the magnitude of the operation amount for the own vehicle, evaluating the state of the own vehicle, evaluating the risk of each labeled object with respect to the own vehicle, and A weighting coefficient to be used to form an evaluation function, and the control amount calculating means includes weighting coefficient changing means for changing the value of the weighting coefficient for the object when the label management means receives a request for adding or deleting a label. It is characterized by the following.
[0006]
【The invention's effect】
ADVANTAGE OF THE INVENTION According to this invention, the driving | operation assistance apparatus which can implement | achieve a smooth control which does not give a driver an uncomfortable feeling can be provided.
[0007]
BEST MODE FOR CARRYING OUT THE INVENTION
Hereinafter, embodiments of the present invention will be described in detail with reference to the drawings. In the drawings described below, those having the same functions are denoted by the same reference numerals, and repeated description thereof will be omitted.
[0008]
First embodiment
A first embodiment of the present invention will be described with reference to FIGS. FIG. 1 is a layout diagram showing the configuration of the entire driving assist system.
In FIG. 1, a front radar 1a is attached to the front of the host vehicle A and measures the positions of a plurality of vehicles located in front of the host vehicle A. The image sensor 1b is also mounted at an appropriate position on the front of the vehicle A, and complements the measurement information of the forward radar 1a. The rear radar 1c is attached to the back of the host vehicle A, and measures the positions of a plurality of vehicles located behind the host vehicle A. The side sensors 1d are attached to the left and right sides of the own vehicle A one by one, and detect the position of the vehicle located on the side of the own vehicle A, which is a blind spot of the front radar 1a and the rear radar 1c. Note that a radar can be used as the side sensor 1d, but an ultrasonic sensor or an image sensor can be used.
[0009]
The vehicle speed sensor 2 can be realized by attaching a rotary encoder to a wheel. A pulse train having a cycle corresponding to the rotation speed of the wheel is output to obtain a measured value of the vehicle speed.
[0010]
An accelerator pedal angle sensor 4 is attached to the accelerator pedal 3, and measures the driver's depression angle of the accelerator pedal 3. Further, a reaction force motor 5 is attached to the accelerator pedal 3, receives a signal from the calculation unit 6, adjusts the magnitude of the reaction force applied to the accelerator pedal 3, and presents a driver with information on a desired accelerator angle. .
[0011]
The arithmetic unit 6 is composed of a microcomputer and its peripheral parts, processes signals from each sensor according to a program recorded in a built-in memory, and controls the reaction force motor 5.
[0012]
The system start / stop switch 7 is attached to an appropriate place in the cabin of the host vehicle A, and starts the system by pressing the start switch to start the driving assist operation. Further, by pressing the stop switch, the operation of the arithmetic unit 6 is switched so that the auxiliary operation is stopped.
[0013]
The arithmetic unit 6 comprises blocks 6a, 6b and 6-1 to 6-4 shown in FIG.
[0014]
In FIG. 2, reference numeral 10 denotes a surrounding state detecting means, which comprises a forward radar 1a, an image sensor 1b, a backward radar 1c, and a side sensor 1d. Reference numeral 11 denotes a vehicle state detecting means, which includes a vehicle speed sensor 2 and an accelerator pedal angle sensor 4. The calculating unit 6 includes an evaluation function forming unit 6a and a control amount calculating unit 6b. The evaluation function constructing means 6a includes a label assigning means 6-1, a label deleting means 6-2, and a label managing means 6-3, and the control amount calculating means 6b includes a weighting coefficient changing means 6-4. Reference numeral 12 denotes an actuator, which comprises the reaction motor 5.
[0015]
Hereinafter, a specific configuration method and a processing procedure of each block will be described taking a road situation as shown in FIG. 3 as an example.
[0016]
In FIG. 3, in a scene where the host vehicle A is traveling on a one-lane road on one side, and the rear vehicle C is following the road, the preceding vehicle traveling forward at a slower speed than the host vehicle A. The scene where B appeared is shown. When assisting driving, it is necessary to recognize an object that restricts the travel of the host vehicle A, and to prompt an operation so as not to approach the object more than necessary. Here, it is assumed that the recognized objects are numbered “1”, “2”, and “3”, and the control amount calculation process is performed. The correspondence between the attached label and the recognized object is recorded and held in the label management unit 6-3.
[0017]
In this case, the objects that may restrict the traveling of the host vehicle A are the rear vehicle C and the preceding vehicle B. The label assigning unit 6-1 recognizes these vehicles and issues a request to the label managing unit 6-3 to assign a label as an object that can be an obstacle to the vehicle A. The label may be provided at the same time when the object enters the detection range of the sensor, but the label provision section may be made smaller than the detection range as shown in FIG. Assuming that the labels are assigned in the order of detection, the rear car C is labeled "1", and when the preceding car B enters the labeling section, the label "2" is assigned to the preceding car B. Attached. Note that the label deletion unit 6-2 is attached to the target object when the detected object goes out of the detection range or when it is no longer a target of the risk evaluation because it is sufficiently separated from the vehicle A. A request is issued to the label management means 6-3 to delete the label. For example, it is assumed that a label holding section is defined as shown in FIG. 3 and a label deletion request is issued when the detected object comes out of the label holding section. If the label holding section is made wider than the labeling section and hysteresis is introduced for label addition and deletion, label addition and deletion will not be repeated even if the detected object stays on the boundary of the label addition section. This has the effect of stabilizing the labeling process.
[0018]
The surrounding situation detecting means 10 detects the distance to each object and the relative speed, and can calculate the absolute speed of each object together with the own vehicle speed detected by the own vehicle state detecting means 11. By determining the position of the host vehicle A as an appropriate value (for example, 0), the position of each object can be defined. Hereinafter, the position and the velocity of the object labeled with the label “i” thus determined are respectively expressed by x_i, V_iWill be expressed as
[0019]
For the labeled object, the risk degree for the vehicle A is evaluated. Although various indicators can be considered as the index of the risk degree, for example, inter-vehicle time (inter-vehicle distance / vehicle speed) can be used as the index. Assuming that the evaluation term is configured so that the smaller the value is, the better the evaluation is, the risk degree evaluation for the object labeled i is as follows.
[0020]
(Equation 1)

Can be used.
[0021]
In order to calculate a desired operation for the vehicle A, a term for evaluating the operation itself for the vehicle A is required as an evaluation function, in addition to an evaluation item for the surrounding environment such as Expression (1). For example, the acceleration / deceleration command value for the vehicle A is set to u_xIf so, the following evaluation terms are required.
[0022]
(Equation 2)

In addition, the state of the host vehicle A can be added to the evaluation. For example, the desired traveling speed of the vehicle A is represented by v^dIs the square of the deviation between the desired vehicle speed and the actual traveling vehicle speed.
[0023]
(Equation 3)

Can be added to the evaluation term. An expression obtained by adding a weighting coefficient to the above-described evaluation terms and taking the sum is introduced as an evaluation expression. That is,
[0024]
(Equation 4)

Is defined as an evaluation amount at a certain moment. Where w_X, W_V, W_iIs a weighting coefficient for each evaluation term. The evaluation function is defined as an amount obtained by integrating the function L over a predetermined evaluation section. That is,
[0025]
(Equation 5)

Is an evaluation function. Here, t represents the current time, and T represents the length of the evaluation section. The above-described processing is performed in the evaluation function construction means 6a, and information on the constructed evaluation function is sent to the control amount calculation means 6b.
[0026]
The control amount calculating means 6b receives the information of the evaluation function set by the evaluation function setting means 6a and calculates a control amount for controlling the actuator. The calculation of the control amount is performed by manipulating the operation amount u for the vehicle A such that the evaluation function (5) is minimized._x(Τ), t ≦ τ ≦ t + T, and the calculated operation amount u_xA command value for controlling the reaction force to urge the driver to perform an operation along (τ) is sent to the reaction force motor 5.
[0027]
In order to evaluate the evaluation function (5), it is necessary to predict the state of the vehicle A and other objects in the evaluation section t ≦ τ ≦ t + T. In the scene of FIG. You can think of things.
Regarding own vehicle A,
[0028]
(Equation 6)

Predicts the state quantity. As a prediction formula of the rear vehicle C, as a model following the own vehicle A,
[0029]
(Equation 7)

Can be used. Where k₁, K₂Is a constant representing the following characteristic of the following vehicle C, h_BIs a target inter-vehicle time of the rear vehicle C. For the preceding vehicle B, it is assumed that the vehicle travels at a constant speed.
[0030]
(Equation 8)

Model is used.
[0031]
As described above, if an equation for predicting the state quantities of the vehicle A and each object during the evaluation section t ≦ τ ≦ t + T can be defined, the problem of finding a control amount that minimizes the equation (5) is optimal. It is formulated as a control problem, and its systematic solution is well studied (for example, Ref. 1: Kanichiro Kato, {Engineering Optimal Control}, Nonlinear Approach, etc.).
[0032]
The problem here is that the preceding vehicle B enters the labeling section, and in the third term on the right side of the evaluation term (4), in addition to the evaluation term of only the rear car C up to that point, the preceding vehicle B This is when an evaluation term is added. In general, in the optimal control problem, if a continuous prediction equation and a continuous evaluation function are used, the optimal solution also changes continuously with respect to a continuous change in the state quantity at the beginning of the evaluation. If the optimum solution changes continuously, the control amount of the actuator 12 generated from the optimum solution also changes continuously, so that the presented reaction force becomes smooth. However, when the evaluation term for the preceding vehicle B is added, since the evaluation function itself changes discontinuously, the obtained optimal solution also exhibits discontinuity reflecting the change, and the presented reaction force also changes discontinuously. Changes to Reference 2: T.I. Ohtsuka, “Continuation / GMRES method for fast algorithm of nonlinear reducinghorizon control Proc. 39th IEEE Conference on Decision and Control, pp. 766-771, @ 2000. When using an algorithm that solves the optimal control problem at high speed based on the continuity of the solution as shown in the above, the discontinuity of the evaluation function is contrary to the assumption of the continuity of the solution. There is a concern that the performance of the device may deteriorate.
[0033]
In order to compensate for such a discontinuity, a weighting coefficient changing means 6-4 is provided in the control amount calculating means 6b. Before the label is given to the preceding vehicle B, the evaluation expression (4) is
[0034]
(Equation 9)

When the label is given to the preceding vehicle B,
[0035]
(Equation 10)

Changes to In this state, as described above, the evaluation term w₂l₂May cause a problem of discontinuity. Therefore, the weighting coefficient w which is originally a fixed parameter₂Is a time-varying parameter that changes with time t. That is,
[0036]
[Equation 11]

And Now, the time when the label is given to the preceding vehicle B is represented by t.₀As w₂(T₀) = 0 so that w₂If (t) is set, time t₀Since the evaluation function (11) matches the evaluation function before the label is assigned to C, the continuity of the evaluation function can be maintained. Time t₀Hereafter, w₂(T) is replaced with the original design value w₂If the value is gradually increased, the evaluation function can be continuously transformed from Expression (9) to Expression (10), so that the resulting optimal solution also changes continuously, and the control amount Also change smoothly.
[0037]
w₂As a way of changing (t), for example, as shown in FIG. 4, the rate of change is kept constant, and the original design value w₂It is conceivable to increase it gradually. This can be expressed by the following equation.
[0038]
(Equation 12)

Where T_tIs w₂This is a parameter for determining the change rate of (t). Generally T_tIs large, it is expected that the evaluation function changes slowly, and that the obtained control amount also changes more smoothly. But T_tIs too long, the evaluation of the newly detected object becomes difficult to be reflected, and in a situation where the newly detected object approaches the own vehicle A suddenly, the response tends to be delayed, and the driver May cause discomfort. Therefore, the parameter T_tIs the collision time T of the newly detected object with respect to the host vehicle A._CA method of adjusting according to the situation can be considered. That is, in the scene of FIG.
[0039]
(Equation 13)

[0040]
[Equation 14]

The following adjustment method can be used. Where T_t ^min, T_t ^maxIs the parameter T_tIs a constant representing the lower and upper limits of the adjustment range of_C ^min, T_C ^maxIs a suitable constant with the dimension of time.
[0041]
As a specific method of calculating the reaction force command value, for example, the following method can be considered. The value at the current time t of the optimal solution that minimizes the evaluation function (5) is expressed as u_x ^*(T), and the acceleration of the vehicle is u_x ^*The depression angle of the accelerator pedal 3 such that (t) is θ^*(T). The depression angle of the accelerator pedal 3 actually operated by the driver is represented by θ (t), and the reaction force characteristic of the normal accelerator pedal 3 to which no reaction force control is applied is represented by F (θ). At this time, the control command is output by the driving force of the reaction force motor so that the actual reaction force matches the following equation.
[0042]
(Equation 15)

Where K is an appropriate positive gain, f is the upper limit of the reaction force correction amount,
[0043]
(Equation 16)

It is. That is, when the actual stepping angle θ (t) is larger than the stepping angle corresponding to the optimal solution, the reaction force is increased to prompt the driver to decelerate, and in the opposite case, the reaction force is reduced. Is notified that the accelerator pedal 3 can still be depressed.
[0044]
FIG. 5 is a flowchart summarizing the above processing.
In step 1, upon receiving a signal from the surrounding situation detecting means 10, an object that can be an obstacle is identified.
In step 2, it is checked whether there is an obstacle newly entering the label provision section. When a new obstacle is detected, an evaluation term and a prediction equation corresponding to the obstacle are generated in steps 3 and 4, respectively, and then the process proceeds to step 5.
In step 5, the weighting coefficient used for solving the optimal control problem is set. The weighting coefficient that has already reached the original design value is kept as it is, the weighting coefficient whose value is gradually increasing is updated to a new value, and the newly added weighting coefficient is set to zero.
In step 6, the optimal control problem is solved to obtain an optimal solution based on the weighting factors set as described above.
In step 7, the command value of the reaction force is calculated according to the equation (15) based on the obtained value of the starting point of the optimal solution, and the processing of one cycle is completed.
[0045]
FIG. 6 shows an example of a result obtained when the present invention is applied to the scene of FIG.
For simplicity, here the driver calculates the optimal solution u_x ^*Considering that (t) is accurately traced, u_x ^*The change of the value of (t) is shown. Before detecting the preceding vehicle B, the desired vehicle speed v^dIt is assumed that the vehicle was traveling at a constant speed. As shown in FIG. 6, when the evaluation for the preceding vehicle B is introduced with the weighting coefficient fixed at the original design value, the optimal solution changes discontinuously at the moment when the evaluation for the preceding vehicle B is entered. are doing. The discontinuous change of the optimal solution causes the discontinuity of the reaction force presented through the reaction force calculation expression of Expression (15). On the other hand, when the weighting coefficient is adjusted as shown in the first embodiment of the present invention, since the optimal solution changes continuously, the presented reaction force also changes continuously. become.
[0046]
As described above, the first embodiment of the present invention includes a surrounding situation detecting unit 10 for detecting the surrounding situation of the host vehicle A, a host vehicle state detecting unit 11 for detecting the state of the host vehicle A, Evaluation function constructing means 6a for constructing an evaluation function for calculating the desirability of the driving operation for the vehicle A based on signals from the detecting means 10 and the vehicle state detecting means 11; and host vehicle equipment for prompting the driver to perform the desired operation. , And an actuator 12 for controlling the own-vehicle device by a signal from the control amount calculating means 6b. The evaluation function forming means 6a includes a signal from the surrounding state detecting means 10 A label assigning unit 6-1 for requesting an object that can be an obstacle to the host vehicle A to assign a label for calculating a risk degree to the object, A label deletion unit 6-2 for requesting that the label be deleted when the risk degree evaluation for the object is terminated; a correspondence relationship between the label and the labeled object is recorded; A section for evaluating the magnitude of the operation amount for the own vehicle A; and a section for evaluating the state of the own vehicle A, including a label managing means 6-3 for updating the correspondence in response to a request from the means 6-2. An evaluation function is composed of a term for evaluating the degree of risk of the object to which the label is attached to the vehicle A and a weighting factor reflecting the importance of the evaluation for each item. When the means 6-3 receives a request for adding or deleting a label, a weight coefficient changing means 6-4 for changing the value of the weight coefficient for the corresponding object is provided. With such a configuration, discontinuous changes in the evaluation function can be compensated for by adjusting the weighting coefficient, so that the calculated control command can be made smooth.
[0047]
In the case of formulating using an evaluation function, the evaluation term corresponding to the obstacle and the weight for the evaluation term are set, but the weight set there has no physical meaning, and the driver needs to approach the obstacle. Since this is a subjective parameter for which an optimal value is determined depending on how much is allowed, in the present embodiment, the arithmetic processing is performed not as an estimated parameter but as a control parameter. In addition, the number of obstacles around the host vehicle that must be considered when driving assistance is not always one, and the number of obstacles changes every moment depending on the driving environment, and with the change in the number of obstacles, The configuration of the evaluation function itself has also changed. On the other hand, in the control of driving assistance, smooth control is required in order not to give a sense of incongruity to the driver.If the evaluation function changes discontinuously, the calculated control command also changes discontinuously. Control is not always smooth. In addition, generally, the optimization calculation for obtaining the operation amount that minimizes the evaluation function often requires a high calculation load, and particularly when the control target and the evaluation function are non-linear, the calculation time becomes a serious problem.
[0048]
For this reason, in the present embodiment, by taking the above configuration, the calculation time is shortened by assuming the continuity of the change of the solution, and the discontinuity of the evaluation function and the continuity of the change of the solution may be lost. As a result, the performance of the algorithm is not adversely affected, and smooth control without discomfort is realized.
[0049]
Further, the labeling means 6-1 requests that a label be given to the target object when the distance between the object that can be an obstacle to the host vehicle A and the host vehicle A becomes smaller than a predetermined value. Then, the label deletion unit 6-2 requests deletion of the label when the distance between the object and the host vehicle A becomes larger than another predetermined value or when the detection by the surrounding situation detection unit 10 becomes impossible. It has become. With such a configuration, it is possible to add hysteresis to the determination of the obstacle to be evaluated, so that even when the obstacle stays on the boundary of the range to be evaluated, the label is added and deleted. Can be prevented from occurring frequently, and a more stable control command can be calculated.
[0050]
When the label management unit 6-3 detects a new label assignment request, the control amount calculation unit 6b sets the value of the weighting coefficient for the risk degree of the object to which the new label is assigned to 0, Thereafter, the value of the weighting coefficient is gradually increased over a predetermined time. With such a configuration, at the moment when a new obstacle needs to be included in the evaluation, the evaluation function is incorporated into the evaluation while maintaining the continuity of the change, so that a new obstacle can be detected. , The change in the control amount can be made smooth.
[0051]
Further, as shown in Expressions 12 and 13, the rate of change of the value of the weighting coefficient in the weighting coefficient changing means is set to a constant value, and the rate of change of the value is determined as the collision time (= inter-vehicle distance / relative speed) with respect to the newly detected object ). With such a configuration, the rate of change of the weighting coefficient is determined according to the collision time with the obstacle, so that the balance between the response speed of the control system and the smoothness of the control can be changed according to the approach speed of the object. it can.
[0052]
Further, the evaluation function forming means 6a is configured to form an evaluation function that includes an acceleration / deceleration command for the vehicle A in the evaluation as the magnitude of the operation amount for the vehicle A to be evaluated. With such a configuration, since the operation amount for the host vehicle A is to be evaluated, it is possible to prevent an excessive control amount from being calculated and calculate the control amount within a realistic operation range.
[0053]
Further, the evaluation function forming means 6a is configured to form an evaluation function that includes the speed of the host vehicle A in the evaluation as the state of the host vehicle A to be evaluated. With such a configuration, since the state of the host vehicle A is to be evaluated, it is possible to calculate a control amount in consideration of the driver's desired traveling vehicle speed, and to assist the driving operation according to the driver's intention.
[0054]
In addition, the evaluation function constructing means 6a sets the risk degree of the labeled object to the own vehicle A as the distance to the own vehicle A is smaller or as the speed of approaching the own vehicle A is larger. An evaluation function is configured by using a term to be greatly evaluated. With such a configuration, the evaluation of the risk degree is determined according to the distance to the obstacle and the relative speed. Therefore, the potential risk degree at which the vehicle A collides with the obstacle is objectively evaluated, and the obstacle is evaluated. It is possible to calculate a control amount that encourages maintaining an appropriate distance from the object.
[0055]
The rate of change of the value of the weighting coefficient in the weighting coefficient changing means 6-4 is set to 0 at the beginning and end of the section in which the value is changed, and the average rate of change of the value is set to the collision time with respect to the newly detected object. By deciding accordingly, it is possible to configure the rate of change of the weighting coefficient to be continuous, and to realize more smooth control than when the value of the weighting coefficient is continuous.
[0056]
Further, the rate of change of the weighting coefficient in the weighting coefficient changing means 6-4 is set to a constant value, and the rate of change of the value is determined according to the relative speed with respect to the corresponding object. An object which is determined according to the relative speed and need not be considered when calculating the control amount can be quickly excluded from the evaluation target.
[0057]
Further, the rate of change of the value of the weighting coefficient in the weighting coefficient changing means 6-4 is set to 0 at the beginning and end of the section in which the value is changed, and the average rate of change of the value is set to the relative speed with respect to the object to be deleted. By deciding accordingly, the rate of change of the weighting coefficient is configured to be continuous, and more smooth control can be realized as compared with the case where the value of the weighting coefficient is continuous.
[0058]
The above is the first embodiment of the present invention.
[0059]
Second embodiment
A second embodiment of the present invention will be described with reference to FIGS.
The main configuration of the second embodiment is the same as that of the first embodiment, and the layout is shown in FIG. 1 and the configuration is shown in FIG.
Hereinafter, a specific operation of each block in the second embodiment will be described by taking a road condition as shown in FIG. 7 as an example.
[0060]
In FIG. 7, the vehicle speed of the rear vehicle C decreases from the state in which the preceding vehicle B, the own vehicle A, and the rear vehicle C are traveling on the one-lane road in one order, and the rear vehicle C becomes an obstacle of the own vehicle A. A scene outside the label holding section, which is an evaluation section, is shown. It is assumed that a label “1” is given to the rear vehicle C and a label “2” is given to the preceding vehicle C. Assuming that the same evaluation expression as that of the first embodiment is used, the evaluation expression is the same as Expression (10).
[0061]
Since the rear vehicle C has moved out of the label holding section, the label deletion unit 6-2 requests the label management unit 6-3 to delete the label “1” assigned to the rear vehicle C. Put out. If you immediately respond to the deletion request, the evaluation formula is
[0062]
[Equation 17]

And the evaluation term w for the rear vehicle C₁l₁, A discontinuity may occur, and the same inconvenience as that described in the first embodiment may occur. Therefore, even if the label management unit 6-3 receives the deletion request, the label management unit 6b does not immediately delete the label "1" of the rear vehicle C, and changes the weighting coefficient corresponding to the label "1" to the control amount calculation unit 6b. Command. That is, the evaluation expression
[0063]
(Equation 18)

As a weighting coefficient w₁(T) is the original design value w₁And when the value decreases to 0, the label “1” is deleted. By doing so, the continuity of the evaluation function can be maintained, and the change in the generated control amount can be made smooth.
[0064]
Weighting coefficient w₁As a way of changing (t), as shown in equation (12) of the first embodiment, a method of keeping the rate of change constant and gradually decreasing it can be used. In this case, since the speed of response to the object to be deleted does not matter so much, it is possible to adopt a method of changing the weighting coefficient with an emphasis on smoothness. Specifically, the weighting coefficient is not only changed continuously but also smoothly changed so that the change rate becomes continuous as shown in FIG. In the form of an expression, for example,
[0065]
[Equation 19]

An expression such as Where t₀Is the time when the label management unit 6-3 receives the request to delete the label "1". When a change like the equation (19) is made,
[0066]
(Equation 20)

Holds, the rate of change becomes 0 at both ends of the section where the weighting coefficient is changed, and not only the value but also the rate of change are continuous, so that a smoother change can be achieved. Such a change expression emphasizing smoothness suppresses a change in the value at the beginning of the change, so that the response becomes slow, but does not pose a problem in a situation where the response is not so important. As for the average response speed, as in the first embodiment, the parameter T_tIt is possible to adjust by changing. However, the scene in which the label is deleted is a scene in which the target object leaves, and the collision time becomes infinite. Therefore, the adjustment method based on the collision time as in Expression (13) does not make sense. . Therefore, a method such as the following formula for adjusting the relative speed instead of the collision time can be considered.
[0067]
(Equation 21)

Where R_d ^max, R_d ^minIs a suitable constant with the dimension of velocity.
[0068]
In a situation in which the object is moving away, there is a possibility that the object goes out of the detection range of the sensor before the weighting coefficient becomes 0, and the evaluation term (1) cannot be calculated. In particular, when the label holding section is set in the same range as the detection range of the sensor, the loss of the corresponding object triggers a label deletion request, so that information on the corresponding object cannot be obtained any more. In such a case, the position and speed of the lost object are estimated using the data immediately before the object is lost, and the expression (1) is evaluated using the estimated value. For example, in the scene of FIG. 7, the time t immediately before the rear car C is lost^blPosition and velocity at x_B ^bl, V_B ^bl, An estimated value is generated according to the following equation.
[0069]
(Equation 22)

FIG. 9 summarizes the above processing in a flowchart.
In step 1, a signal from the surrounding situation detecting means 10 is received, and a potential obstacle is identified.
In step 2, it is checked whether there is an obstacle that has come out of the label holding section.
If there is a corresponding object, a label deletion request is issued to the object in step 3. At the stage of step 3, only a deletion request is issued, and deletion is not performed immediately.
In step 4, the presence / absence of a lost object is checked by the sensor among the objects to which the label has been given.
If there is, in step 5, an estimated value as shown in equation (22) is generated, and the estimated value is used instead of the detected value.
In step 6, the weighting coefficient is updated. The weight of the object for which the label deletion request has been issued is reduced, and the weight of the other object is retained.
In step 7, the value of the updated weighting coefficient is checked, and it is checked whether or not any of the weighting coefficients has become 0.
If any value has become 0, the corresponding label, the evaluation term and the prediction formula are deleted in step 8.
In step 9, the optimal control problem is solved by using the evaluation function and the prediction formula composed of the remaining evaluation terms and the prediction formula, and the optimum solution is obtained.
In step 10, the command value of the reaction force is calculated according to the equation (15) based on the obtained value of the starting point of the optimal solution, and the processing of one cycle is completed.
[0070]
FIG. 10 shows an example of a result obtained when the present invention is applied to the scene of FIG.
As in the first embodiment, here the driver determines the calculated optimal solution u._x ^*Considering that (t) is accurately traced, u_x ^*The change of the value of (t) is shown. From time 0 to time t^bdUntil the three vehicles are traveling at a constant speed at the same speed, and at time t^bdAt the time t₀It is assumed that the rear car C comes out of the label holding section. When the inter-vehicle distance with the following vehicle C increases, the inter-vehicle distance with the preceding vehicle C increases, so that the optimal solution is a solution in which a slight deceleration is commanded and then the vehicle is accelerated again. FIG. 10 shows a case where the weighting factor is fixed, a case where the weighting factor is gradually reduced at a constant rate of change similarly to the equation (12), and a case where the weighting factor is continuously changed to the rate of change as shown in the equation (19). 3 shows three results in the case of gradual decrease with the property. When the weighting factor is fixed, as in the first embodiment, the time t₀In u_x ^*The value of (t) changes discontinuously. On the other hand, when the rate of change is constant and the weight is gradually reduced, the time t₀U in_x ^*Although the continuity of (t) is maintained, there is a slight gap in the rate of change. When the rate of change of the weighting coefficient is given continuity, u_x ^*(T) also changes smoothly.
[0071]
As described above, in the second embodiment of the present invention, when the label management unit 6-3 detects a label deletion request, the control amount calculation unit 6b allocates the label to be deleted. The value of the weighting coefficient for the risk degree of the object is gradually reduced to 0 over a predetermined time, and after a predetermined time, the corresponding label is deleted. With such a configuration, when the detected object is excluded from the evaluation of the control amount calculation, it is excluded from the evaluation while maintaining the continuity of the change of the evaluation function. It can be smooth.
[0072]
Further, as shown in Expressions 21 and 22, the control amount calculating means 6b cannot detect the target object if the surrounding situation detecting means 10 cannot detect the target object before the value of the weighting coefficient becomes zero. The control amount is calculated using a provisional value extrapolated from the distance to the object and the relative speed at the time immediately before. With such a configuration, not only the data directly detected from the surrounding situation detecting means 10 but also a provisional value is used as necessary, so that the influence of the lost obstacle on the control amount calculation can be reduced. .
[0073]
The above is the second embodiment of the present invention.
[0074]
Third embodiment
A third embodiment of the present invention will be described with reference to FIGS.
The main configuration of the third embodiment is the same as that of the first embodiment, and the layout is shown in FIG. 1 and the configuration is shown in FIG.
Hereinafter, a specific operation of each block in the third embodiment will be described by taking a road situation as shown in FIG. 11 as an example.
[0075]
FIG. 11 shows a state in which the vehicle is traveling on a one-lane road on one side at substantially the same speed in the order of the preceding vehicle B, the host vehicle A, and the rear vehicle C. The host vehicle A is running slightly closer to the preceding vehicle B. It is assumed that both the preceding vehicle B and the following vehicle C are traveling inside the labeling section in which the own vehicle A evaluates as an obstacle. Here, in such a situation, a situation where the system start switch of the driving assist device is pressed to start the driving assist operation will be particularly considered. The moment when the start switch is pressed is defined as time t = 0.
[0076]
When the system is started, the surrounding situation detecting means 10 detects the preceding vehicle B and the following vehicle C inside the labeling section. When the result is sent to the arithmetic unit 6, the label assigning means 6-1 immediately requests the label managing means 6-3 to assign labels to both vehicles. Upon receiving the request, the label management unit 6-3 assigns an appropriate label to each vehicle. Here, it is assumed that the label “1” is given to the preceding vehicle B and the label “2” is given to the following vehicle C.
[0077]
The evaluation function structuring means 6a receives the detection result, forms the same evaluation expression as the expression (10), and passes the processing to the control amount calculating means 6b. The control amount calculating means 6b corrects the weighting coefficients. In this case, since both labels 1 and 2 are new detection targets, the evaluation formula is reconfigured so that both weighting coefficients are variable. That is,
[0078]
[Equation 23]

And the evaluation formula are set. For changing the weighting coefficient, a method according to the equations (12) to (14) in the first embodiment can be used. Further, the equations (6) to (8) in the first embodiment can be applied to the equation for predicting the behavior of each vehicle.
[0079]
The difference between the operation immediately after the system startup and the other time is that immediately after the system startup, the information on the optimal solution generated before that time cannot be used. This means that when solving an optimization problem, it is impossible to use a method that efficiently solves the problem using the solution obtained immediately before, which increases the computational load and the value of the solution obtained first. However, there is a problem that the presented reaction force changes discontinuously, giving the occupant a sense of incongruity.
[0080]
As a method for solving such a problem, Literature 2 makes the length T of the evaluation section in the evaluation function (5) variable, sets the length of the evaluation section to 0 at time t = 0, and then sets time A method is disclosed in which the length of the evaluation section is gradually increased to the original design value as time elapses. When the length of the evaluation section is set to 0, a trivial solution in which the operation amount is always 0 is obtained as the optimal solution. Since calculation is practically unnecessary to obtain an obvious solution, the problem of calculation load is avoided. Further, since the obtained operation amount is 0, a command for correcting the reaction force is not issued immediately after the system is started, which also solves the problem of discontinuity.
[0081]
As a setting method of the length of the evaluation section, the following equation is given as an example.
[0082]
(Equation 24)

Here, T (t) is the length of the evaluation section at time t,₀Is an original design value of the evaluation section length, and α is an appropriate positive constant.
[0083]
By using the adjustment of the evaluation section length and the adjustment of the weighting coefficient as described above, it is possible to smoothly change the control amount from the start of the system to the steady operation.
[0084]
FIG. 12 shows an example of a result obtained immediately after starting the system in the scene of FIG.
Here, as in the first embodiment, u_x ^*The change of the value of (t) is shown. In addition, when both the evaluation section length and the weighting coefficient are fixed, the evaluation section length is variable, but when the weighting coefficient is fixed, the evaluation section length and the weighting coefficient are both variable. Is shown.
[0085]
In the scene shown in FIG. 11, since the host vehicle A is too close to the preceding vehicle B, a command prompting deceleration should be output to increase the inter-vehicle distance. When both the evaluation section and the weighting coefficient are fixed, a deceleration command is generated immediately after the system is started, and a large auxiliary reaction force is generated. When the evaluation section is variable, the initial optimal solution is 0, and no auxiliary reaction force occurs at time t = 0. When the weighting coefficient is fixed, the rate of change of the optimal solution at time t = 0 remains discontinuous. However, when the weighting coefficient is also variable, the rate of change of the optimal solution becomes continuous, and a smooth reaction force is obtained. A presentation can be made.
[0086]
The above is the third embodiment of the present invention.
[0087]
Although the present invention has been specifically described based on the embodiments, the present invention is not limited to the above-described embodiments, and it is needless to say that various modifications can be made without departing from the gist of the present invention.
[Brief description of the drawings]
FIG. 1 is a diagram showing an arrangement diagram of a first embodiment of the present invention.
FIG. 2 is a diagram showing a configuration diagram of a first embodiment of the present invention.
FIG. 3 is a diagram illustrating an example of an application scene of the first embodiment of the present invention.
FIG. 4 is a diagram illustrating an example of a method of changing a weighting coefficient according to the first embodiment of the present invention.
FIG. 5 is a flowchart illustrating a procedure of a process according to the first embodiment of the present invention.
FIG. 6 is a diagram illustrating an example of a result obtained when the first embodiment of the present invention is applied.
FIG. 7 is a diagram illustrating an example of an application scene of a second embodiment of the present invention.
FIG. 8 is a diagram illustrating an example of a method of changing a weighting coefficient according to the second embodiment of this invention.
FIG. 9 is a flowchart illustrating a procedure of a process according to the second embodiment of the present invention.
FIG. 10 is a diagram illustrating an example of a result obtained when the second embodiment of the present invention is applied.
FIG. 11 is a diagram illustrating an example of an application scene of a third embodiment of the present invention.
FIG. 12 is a diagram showing an example of a result obtained when the third embodiment of the present invention is applied.
[Explanation of symbols]
1a ... Forward radar
1b ... Image sensor
1c ... rear radar
1d ... side sensor
2: Vehicle speed sensor
3. Accelerator pedal
4: Accelerator pedal angle sensor
5: Reaction motor
6 arithmetic unit
6a ... Evaluation function construction means
6-1 ... Labeling means
6-2 ... Label deletion means
6-3 ... Label management means
6b: Control amount calculating means
6-4: Weighting coefficient changing means
7. System start / stop switch
A: Own vehicle
10: Surrounding situation detecting means
11 ... own vehicle state detecting means
12. Actuator

Claims (12)

An ambient condition detecting means for detecting an ambient condition of the vehicle;
Own vehicle state detecting means for detecting the state of the own vehicle,
Evaluation function configuration means for configuring an evaluation function for calculating desirability of a driving operation on the own vehicle by a signal from the surrounding situation detection means and the own vehicle state detection means,
Control amount calculating means for calculating a control signal for the own vehicle device to prompt the driver to perform a desired operation;
An actuator that controls the host vehicle device based on a signal from the control amount calculation unit,
The evaluation function configuration means,
A label providing unit that requests to give a label for calculating a risk degree for the object to an object that may become an obstacle to the own vehicle by a signal from the surrounding situation detecting unit,
Label deletion means requesting to delete the label when discontinuing the risk assessment for the object given the label,
A label management unit that records a correspondence relationship between the label and the object to which the label is attached, and updates the correspondence relationship in response to a request from the label assignment unit and the label deletion unit.
A term for evaluating the magnitude of the operation amount for the vehicle,
A term for evaluating the state of the vehicle,
A term for evaluating the risk degree for the subject vehicle of each of the objects provided with the label,
A weighting coefficient that reflects the importance of the evaluation for each of the above terms, and constitutes an evaluation function,
The control amount calculating unit includes a weighting factor changing unit that changes a value of the weighting factor for the corresponding object when the label managing unit receives a request for adding or deleting the label. Driving operation assist device. The label applying means requests that the label be applied to the target object when a distance between the object and the host vehicle that can be an obstacle to the host vehicle becomes smaller than a predetermined value. ,
The label deletion unit requests deletion of the label when the distance between the object and the host vehicle is larger than another predetermined value or when the detection by the surrounding situation detection unit becomes impossible. The driving operation assisting device according to claim 1, wherein The control amount calculating means, when the label management means detects a new label assignment request, sets the value of the weighting coefficient for the risk degree of the object to which the label is newly assigned to 0,
The driving operation assisting device according to claim 1, wherein the value of the weighting coefficient is gradually increased over a predetermined time thereafter. The rate of change of the value of the weighting coefficient in the weighting coefficient changing means is a fixed value, and the rate of change of the value is determined according to the collision time (= inter-vehicle distance / relative speed) with respect to the newly detected object. The driving operation assisting device according to claim 3, wherein The rate of change of the value of the weighting coefficient in the weighting coefficient changing means is set to 0 at the beginning and end of the section in which the value is changed, and the average rate of change of the value is determined according to the collision time with the newly detected object. The driving operation assisting device according to claim 3, wherein the driving operation assisting device is determined. The control amount calculation means, when the label management means detects the label deletion request, the value of the weighting coefficient for the risk degree of the object to which the label targeted for deletion is allocated over a predetermined time The driving operation assisting device according to claim 1 or 2, wherein the label is gradually reduced to 0, and the label is deleted after a predetermined time. The control amount calculating means is configured such that, if the target object cannot be detected by the surrounding situation detecting means before the value of the weighting coefficient becomes 0, the distance to the object at a time immediately before the detection becomes impossible. 7. The driving assist system according to claim 6, wherein the control amount is calculated using a provisional value extrapolated from the relative speed and the relative speed. 8. The driving operation assistance according to claim 6, wherein the rate of change of the value of the weighting coefficient in the weighting coefficient changing unit is a constant value, and the rate of change of the value is determined according to a relative speed with respect to the corresponding object. apparatus. The rate of change of the value of the weighting coefficient in the weighting coefficient changing means is set to 0 at the beginning and end of the section in which the value is changed, and the average rate of change of the value is determined according to the relative speed to the object to be deleted. 8. The driving operation assisting device according to claim 6, wherein the driving operation assisting device is determined. 4. The evaluation function configuration unit according to claim 1, wherein the evaluation function is configured to include an acceleration / deceleration command for the host vehicle in the evaluation as the magnitude of the operation amount for the host vehicle to be evaluated. 10. The driving operation assisting device according to any one of items 9 to 9. The driving according to any one of claims 1 to 10, wherein the evaluation function configuration unit configures an evaluation function that includes a speed of the host vehicle in an evaluation as a state of the host vehicle to be evaluated. Operation assistance device. The evaluation function constructing means, as the risk degree of the object given the label to the own vehicle, the smaller the distance to the own vehicle, or the greater the speed of approaching the own vehicle, the greater the risk degree The driving operation assisting device according to any one of claims 1 to 11, wherein the evaluation function is configured by using a term that greatly evaluates the following.

Images